CHAPTER 1 - Introduction

1.1 Introduction

The aviation industry's dynamic and highly volatile pricing structure is shaped by fluctuating
demand, fuel price volatility, competitive pressures, and complex route logistics. These
variables make flight price prediction a challenging task for airlines, travel agents, and
consumers. Even minor changes can significantly impact ticket prices, affecting consumers
seeking the best deals and airlines striving to maximize revenue and optimize seat allocations.
With the industry's growth, accurate flight price forecasting is increasingly crucial.
Data-driven decision-making is vital across the sector, making real-time flight price
predictions essential for stakeholders. Whether optimizing pricing strategies, offering the best
deals, or timing ticket purchases, predictive models are central to travel planning, resource
allocation, and revenue management. Precision in price prediction is increasingly necessary
as the aviation market becomes more competitive, significantly impacting business strategy
efficiency.

1.2 Challenges with Traditional Price Prediction Models

Traditional methods for predicting flight prices, such as rule-based systems and statistical
models, have consistently fallen short in adapting to the swift and unpredictable changes in
the aviation market. These methods often depend on static variables and historical trends,
failing to account for real-time fluctuations and intricate market conditions. As a result, they
struggle to capture the complex pricing patterns influenced by factors like fuel prices,
macroeconomic trends, seasonal travel demand, competitor actions, and route-specific
logistics. This leads to suboptimal projections, causing airlines to miss revenue opportunities
and consumers to poorly time their ticket purchases.

Moreover, these conventional methods falter when faced with the nonlinear behaviors
inherent in airfare pricing. For instance, an airline might hike prices in response to rising fuel
costs but later reduce them to stimulate demand on underbooked flights. Without the ability
to continuously learn from new data, static models are inadequate for capturing these rapid,
unpredictable shifts. This gap in forecasting accuracy underscores the need for more
advanced approaches that can adapt to dynamic pricing environments in real time.

1.3 Advancements in Artificial Intelligence and Deep Learning

Advancements in artificial intelligence (AI), particularly Generative AI (GenAI) and deep

learning, present promising solutions to the challenges of volatile airfare pricing. AI models
excel at analyzing and processing vast historical data, identifying patterns too complex for
traditional methods. Leveraging deep learning architectures, these systems learn from various

7
market factors, including real-time competitor pricing, seasonal demand shifts, and external
economic events, to generate more accurate and dynamic price forecasts.

Generative AI is particularly notable for generating realistic predictions by modeling the

complex interplay of multiple factors. GenAI systems simulate future pricing scenarios based
on historical data while incorporating real-time changes, allowing airlines and travel agencies
to make informed decisions by anticipating price shifts. When combined with deep learning
techniques like Recurrent Neural Networks (RNNs) and Long Short-Term Memory (LSTM)
networks, AI models become even more powerful. These architectures are designed to
process sequential and temporal data, crucial for capturing evolving pricing trends.

1.4 The Role of LSTM and RNN Architectures

LSTM and RNN architectures are particularly well-suited for the task of flight price
prediction because they are adept at handling time-series data. These models have the unique
ability to "remember" patterns from previous pricing cycles and apply them to future
predictions. LSTMs, in particular, are excellent at retaining long-term dependencies, making
them ideal for analyzing price trends that span weeks or even months. On the other hand,
RNNs are highly effective at capturing short-term fluctuations in pricing data, such as
last-minute fare changes that might occur due to sudden shifts in demand or supply.

By combining LSTM and RNN architectures, the AI system can process both long-term
trends and short-term variations, allowing it to continuously adapt to evolving market
conditions. This adaptability is especially critical in the aviation industry, where prices can
fluctuate multiple times a day depending on factors like booking window and remaining seat
availability. By integrating these architectures, the AI system can generate highly accurate,
real-time forecasts that reflect the most current market conditions.

1.5 Importance of Real-Time Forecasting

One of the most notable advancements in AI-driven systems is their ability to forecast in real
time. Traditional models, which rely on static historical data, often struggle to adapt to
sudden market shifts. In contrast, AI models with real-time forecasting capabilities are
constantly updated with live data streams, allowing for on-the-fly prediction adjustments.
This is invaluable in an industry where prices can change within minutes due to factors like
competitive pricing or last-minute seat availability.

For airlines, real-time forecasting enables dynamic pricing strategies that maximize revenue
by adjusting prices in response to real-time demand. For consumers, it means the chance to
secure the best deals with timely notifications when fares are expected to rise or drop. Travel

8
agencies can use these systems to offer personalized pricing recommendations, enhancing
customer satisfaction and boosting booking rates.

1.6 Motivation and Approach

The exponential rise in air travel is palpable. I've clocked more flights in the last two years
than in the previous eighteen! Stumbling upon my primary research paper on Flight Price
Prediction piqued my curiosity and drove me to delve deeper into the intricate process of
ticket pricing, the myriad factors influencing it, and how AI can play a transformative role.

My approach was straightforward: identify the key factors affecting airline pricing, compile
various pricing data, conduct a literature survey on existing prediction mechanisms, and
uncover the gaps that present opportunities for improvement.

1.7 Objective

This report's primary aim is to delve into the use of Generative AI (GenAI) to enhance flight
price prediction accuracy, highlighting the integration of deep learning models like LSTM
and RNN. It examines how these sophisticated prediction techniques tackle the unique
challenges of volatile airfare pricing. By utilizing historical data, real-time forecasting, and
adaptable AI frameworks, the proposed system strives to deliver precise and timely airfare
predictions. These insights are intended to benefit airlines, travel agencies, and consumers by
refining decision-making processes, optimizing pricing strategies, and helping customers
secure the best deals. The remainder of the report will cover the AI-based system's
methodology, application results, and implications for the aviation industry, providing a
thorough look at how AI can revolutionize flight price prediction in a fiercely competitive
and dynamic market.

9
CHAPTER 2 - Literature Survey

The aviation industry's dynamic and volatile nature has fueled growing interest in applying
Artificial Intelligence (AI) and Machine Learning (ML) to flight price prediction. Traditional
airfare prediction methods fall short in capturing the complex pricing patterns shaped by
fluctuating demand, fuel prices, competitor actions, and route-specific logistics.
Consequently, researchers are turning to advanced AI techniques, including Generative AI
(GenAI), Deep Learning, Recurrent Neural Networks (RNNs), and Long Short-Term
Memory (LSTM) networks, to enhance prediction accuracy in real-time environments. This
literature survey examines the key studies in the field, highlighting their methodologies,
findings, and implications for the future of flight price prediction.

2.1 Generative AI for Real-Time Flight Price Prediction

In Guan's (2024) study, the researcher introduces a Generative AI-based web platform for
real-time airfare prediction. This platform utilizes deep learning models such as RNNs and
LSTMs to process temporal data. By leveraging GenAI, the system can predict future price
trends with high accuracy, benefiting consumers, airlines, and travel agencies alike[4,6]. The
integration of attention mechanisms and real-time forecasting capabilities allows the system
to dynamically update predictions based on current market trends, making it especially
effective in addressing the volatility of the aviation industry.

2.2 Hybrid Models for Time-Series Forecasting

Yang et al. (2022) present a hybrid model for gasoline price forecasting, combining Lasso
regression and Convolutional Neural Networks (CNN) to enhance prediction accuracy[2,3].
Although the study focuses on gasoline prices, the model's capacity to process time-series
data using Lasso regression for variable selection and CNNs for pattern recognition is
adaptable for flight price forecasting. This blend of methods offers a strong framework for
predicting price fluctuations in volatile markets, showcasing the potential of hybrid models
within the aviation sector.

2.3 Energy Forecasting and Its Implications for Airfare Prediction

Hong et al. (2020) provide an extensive review of forecasting techniques in the energy sector,
emphasizing the success of probabilistic models and ensemble methods. Though centered on
energy markets, the parallels with the airline industry's dynamic pricing are evident. By
integrating probabilistic forecasting with AI, future airfare prediction systems could more
effectively manage market volatility and the myriad factors impacting prices, including
external economic shifts and fuel costs.

10
2.4 Handling Price Spikes in Dynamic Environments

Zhang et al. (2023) tackle electricity market price spikes using frequency analysis and
oversampling techniques to boost prediction accuracy[11]. For airfare, where prices can surge
during peak travel seasons or holidays, these methods could enhance a model's capability to
predict and adapt to sudden price increases. This approach could refine real-time flight price
forecasting systems, ensuring they effectively respond to rare but significant price
fluctuations.

2.5 Optimal Flight Ticket Purchase Timing Using Learning to Rank Methods

Oglakcioglu et al. (2023) examine Learning to Rank (LTR) methods to pinpoint the best time
to buy flight tickets. Their research highlights that variables like the number of days until
departure and route-specific behaviors are crucial for predicting airfare. This finding supports
the notion that flight price prediction models must account for dynamic variables that
frequently change as the departure date nears. Therefore, LTR methods can be valuable for
optimizing consumer decisions about when to purchase tickets based on specific route
conditions.

2.6 Holistic Approaches to Airfare Prediction Using Machine Learning

Kalampokas et al. (2023) propose a comprehensive approach to airfare prediction by

integrating Machine Learning (ML), Deep Learning (DL), and Quantum Machine Learning
(QML) techniques to enhance accuracy. They underscore the necessity of using diverse AI
techniques to capture the myriad factors affecting flight prices, such as airline strategies,
external market conditions, and seasonal demand. This approach suggests that future airfare
prediction models could be significantly improved by combining multiple AI frameworks,
thereby better managing the complexities of dynamic pricing environments.

2.7 Survey on Flight Pricing Prediction Using Machine Learning

Rajankar et al. (2019) surveyed various machine learning algorithms for airfare prediction,
such as Support Vector Machines (SVM), Naive Bayes, Decision Trees, and Linear
Regression. They found that while some models, like SVM, excel in static environments,
they falter with the dynamic nature of real-time airfare prediction. This study highlights the
necessity for more advanced models capable of processing real-time data and continuously
updating predictions.

11
2.8 Framework for Predicting Airfare Prices Using ML

Fadhil et al. (2022) propose a framework that uses Decision Trees and K-Nearest Neighbors
(KNN) to predict airfare prices in the Indian airline market. Their findings indicate that
Decision Trees excel at capturing the non-linear relationships between variables like
departure date, route competitiveness, and airline pricing strategies. This study highlights the
significance of non-linear models in forecasting flight prices in volatile markets.

2.9 Flight Fare Prediction Using XGBoost

Sarao and Samanta (2022) compare various ML algorithms in their study and find that
XGBoost delivers the highest accuracy for flight fare prediction. XGBoost's capacity to
manage large-scale datasets and capture complex variable interactions makes it an ideal
choice for real-time price prediction systems. The success of this model suggests that
ensemble learning methods like XGBoost are highly effective in dynamic pricing
environments, such as the airline industry.

2.10 Optimal Purchase Timing Models for Airline Tickets

Noyon et al. (2023) investigate using regression-based models to predict the optimal timing
for purchasing airline tickets. Their study underscores the significance of factors like source
and destination, number of stops, and airline-specific behaviors in determining the best time
to buy. Integrating time-sensitive variables into AI models is crucial for making accurate
predictions in real-time dynamic environments.

2.11 Airfare Analysis and Prediction Using Data Mining

Chawla and Kaur (2017) utilize data mining techniques to investigate the factors affecting
airfare pricing, including oil prices, the number of competitors, and days until departure.
Their comparative analysis of various machine learning algorithms underscores the
significance of historical data mining for accurate price forecasting. This indicates that
integrating data from multiple sources could enhance future models' effectiveness.

2.12 Flight Fare Prediction System Using ML

Finally, Bhosale et al. (2022) present a flight fare prediction system that applies Random
Forest, Decision Trees, and Linear Regression to forecast ticket prices in the Indian domestic
airline market​[12-14]. Their work underscores the potential of ensemble methods and
emphasizes the need for more real-time data to improve model accuracy.

12
 2.13 Research Paper Analysis

Res. Price Deep Neural Gen Merits Demerits Solutions

Paper Prediction Learning Network AI
No.

1 ✔ ✔ ✔ Real-time Computatio Leveraging cloud

airfare nally computing to
forecasting, expensive manage
high accuracy processing costs

2 ✔ ✔ ✔ Hybrid model Model Simplified

improves complexity pre-processing of
accuracy for time series data
gasoline price
prediction

3 ✔ ✔ ✔ Comprehensiv Difficulties Hybrid

e energy in handling approaches using
forecasting very volatile LSTM and CNN
using multiple data for better
sources volatility
handling

4 ✔ ✔ ✔ Short-term Struggles Incorporating

price spikes with external
captured well longer-term macroeconomic
with frequency forecasting indicators for
analysis long-term
forecasting

5 ✔ ✔ ✔ Accurate May not Fine-tuning

timing for generalize algorithms for
purchasing well across specific airline
flights using different routes
Learning to airlines
Rank methods

6 ✔ ✔ ✔ Holistic High model Implementing

machine training time efficient training
learning due to deep mechanisms
approach that learning
accounts for complexity

13
 multiple
factors

7 ✔ ✔ Survey Lacks Applying

provides a implementat machine learning
good ion of models to real
foundation for advanced data
understanding algorithms
flight pricing

8 ✔ ✔ ✔ Framework for Limited to Utilizing

airfare historical generative models
prediction is data and for real-time data
detailed and may not adaptation
effective adapt well to
real-time

9 ✔ ✔ ✔ Machine Predictive Incorporating

learning accuracy adaptive learning
effectively reduces with for changing
predicts fares drastic market conditions
based on market
historical changes
trends

10 ✔ Focus on the Limited Extending

importance of details on analysis to
machine model include more
learning in application modern machine
pricing learning models

11 ✔ ✔ ✔ Combines data Data quality Use of data

mining with affects augmentation
machine model techniques
learning for performance
accurate
predictions

12 ✔ ✔ ✔ ✔ Clear benefits High Employing

from using computation distributed
neural al demand computing to
networks for lower costs
fare prediction

14
13 ✔ ✔ Overview of Does not Extend
machine focus application to
learning specifically specific domains,
algorithms in on price such as flight
various prediction pricing
contexts

14 ✔ ✔ ✔ Deep learning Not focused Could combine

algorithms on price delay predictions
predict flight prediction with fare
delays predictions
effectively

15 ✔ ✔ ✔ ✔ CNN-BiLSTM High Simplifying

hybrid model complexity network
improves architectures
accuracy for
stock
prediction

16 ✔ ✔ New method Requires Automating

enhances significant feature extraction
energy market feature with deep
price engineering learning
predictions

17 ✔ ✔ ✔ Good Highly Using synthetic

application of dependent data for training
deep learning on when real data is
in financial high-quality limited
forecasting data

18 ✔ Multi-source Complexity Streamlining data

urban data in managing pipelines
improves multi-source
dynamic price data
prediction

19 ✔ ✔ ✔ Public Focused on Adapting

transport public methods to air
pricing transport, transportation
strategies are less relevant
well-optimized for flight
prediction

15
20 ✔ ✔ Dynamic price Difficult to Incorporating
prediction for adapt to market volatility
ride-on-deman highly indicators
d services volatile
markets

21 ✔ ✔ Forecasts Not focused Extend maritime

maritime on flight methods to air
transportation prices transportation
demand
effectively

22 ✔ ✔ ✔ LSTM neural Requires Use data

networks extensive augmentation or
provide strong training data generative models
price for training
forecasting
abilities

23 ✔ ✔ ✔ Local housing Limited Generalize

prices scalability to models for
predicted with other broader market
LSTM markets use

24 ✔ ✔ ✔ Cooperative Not focused Integrate price

transport on price forecasting with
systems forecasting transport time
enhanced with predictions
deep learning

25 ✔ ✔ ✔ Strong deep Not Could adapt

learning applicable to methods for fare
techniques for price predictions
travel time prediction
predictions

26 ✔ ✔ ✔ Framework Limited Generalize

effectively scalability framework to
uses ML for across accommodate
airfare price different more airlines
prediction airlines

16
27 ✔ ✔ Strong ML Not focused Apply methods to
techniques for on price pricing
emission prediction predictions
prediction in
transportation

28 ✔ ✔ ✔ ✔ Accurate Focused on Could adapt

predictions rail transit, methods to
using deep not directly forecast flight
learning for applicable to prices by
passenger flow flight incorporating
in urban rail pricing. similar deep
systems. learning
techniques.

29 ✔ ✔ ✔ Machine Limited to Adapt machine

learning public learning methods
provides transportatio to capture pricing
accurate cost n, not easily trends in airline
predictions for generalizabl sectors.
public bus e to airfares.
services.

30 ✔ ✔ ✔ Strong Limited Transfer the

data-driven applicability data-driven
methods for to the airline approach to
price industry. dynamic flight
forecasting, pricing models.
particularly in
energy sectors.

31 ✔ ✔ ✔ Deep learning Not focused Methods could be

accurately on airfare applied to
predicts prediction; forecasting flight
day-to-day mainly pricing with
variations in applicable to proper data
travel demand. large-scale adjustments.
transportatio
n networks.

32 ✔ ✔ ✔ Deep learning Not specific Similar methods

applied for to flight can be adapted
large-scale prices. for airline fare
transport prediction.
networks.

17
33 ✔ ✔ ✔ Reinforcement Focuses on Reinforcement
learning for ride-on-dem learning could be
dynamic price and services. adapted to airfare
prediction. prediction.

34 ✔ ✔ ✔ Prediction of Highly Simplifying the

transportation complex model could
costs using approach. make it
fuzzy analytic applicable to
and neural airfares.
networks.

35 ✔ ✔ ✔ Neural Not focused Adapt energy

network-based on price demand models
demand prediction. to airline pricing.
prediction for
energy in
transport.

36 ✔ ✔ Variable Limited Could focus on

impact application flight prices by
analysis using to airlines. modifying the
neural model.
networks in
transport.

37 ✔ ✔ Neural Focus on Extend the model

network financial to consider flight
analysis of markets price prediction.
share prices rather than
for flight prices.
transportation
companies.

38 ✔ ✔ ✔ Application of Does not Apply

neural focus on sustainability-rela
networks in airfare ted insights to
sustainable prediction. airfares.
transport
systems.

18
39 ✔ ✔ ✔ Route pricing Limited Simplify the
optimization scalability to model for broader
using neural airfares. use in airline
networks. pricing.

40 ✔ ✔ Comprehensiv No specific Extend the review

e review of focus on to include airfare
neural price models.
networks in prediction.
transportation.

41 ✔ ✔ Comparison of Not focused Apply

neural network on price comparison to
frameworks prediction. flight pricing
for public models.
transport.

42 ✔ ✔ ConvLSTM2D Limited Adapt

architecture focus on ConvLSTM2D
for air cargo passenger for flight price
demand airfares. prediction.
forecasting.

43 ✔ ✔ ✔ Graph neural No price Use graph neural

networks for forecasting. networks for
public transit flight price
demand forecasting.
prediction.

44 ✔ ✔ Long-term Focuses on Extend CNN to

prediction energy long-term flight
using CNN predictions. price forecasting.
and
optimizers.

45 ✔ ✔ Neural Focuses on Adapt for

network for cargo, not passenger airfare
Australia's air passenger prediction.
cargo demand. prices.

46 ✔ ✔ ✔ Backpropagati Focuses on Incorporate

on neural demand, not pricing alongside
networks for prices. demand
air passenger predictions.
demand.

19
47 ✔ ✔ ✔ ✔ Neural Focus on Incorporate fare
network for air fleet, not prediction into
cargo fleet fare the model.
assignment. prediction.

48 ✔ ✔ ✔ Forecasting No focus on Combine demand

demand for price forecasting with
low-cost prediction. price prediction
carriers using models.
neural
networks.

49 ✔ ✔ ✔ Forecast of Lacks Incorporate

Australia's pricing pricing alongside
domestic air integration. demand
travel demand. forecasting.

50 ✔ ✔ ✔ ✔ Flight delay No focus on Integrate delay

prediction price and pricing
using deep prediction. prediction for
learning. more robust
models.

51 ✔ ✔ Comprehensiv Focuses on Adapt ML

e review of freight, not methods for
machine passenger passenger airfare
learning for airfares. forecasting.
freight
management.

48 ✔ ✔ Forecasting No focus on Combine demand

demand for price forecasting with
low-cost prediction. price prediction
carriers using models.
neural
networks.

49 ✔ ✔ Forecast of Lacks Incorporate

Australia's pricing pricing alongside
domestic air integration. demand
travel demand. forecasting.

20
50 ✔ ✔ ✔ ✔ Flight delay No focus on Integrate delay
prediction price and pricing
using deep prediction. prediction for
learning. more robust
models.

51 ✔ ✔ ✔ Comprehensiv Focuses on Adapt ML

e review of freight, not methods for
machine passenger passenger airfare
learning for airfares. forecasting.
freight
management.

2.14 Conclusion of Literature Survey

The reviewed literature lays a robust foundation for understanding the potential and
challenges of AI-based flight price prediction systems. The integration of Generative AI,
deep learning, and real-time forecasting techniques shows great promise in improving
prediction accuracy within dynamic environments. However, significant challenges persist,
including data availability, model adaptability, and computational efficiency. Future research
must focus on adaptive learning systems, real-time data integration, and developing more
efficient AI models capable of managing the complexities of dynamic airfare pricing. With
continued advancement, these systems could revolutionize the industry, ensuring accurate,
timely predictions and optimizing decision-making for airlines, travel agencies, and
consumers.

21
CHAPTER 3 - Methodology

Following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses
(PRISMA) guidelines, the methodology for this report encompasses three key steps: the
search strategy, inclusion criteria, and data extraction and analysis. This systematic approach
guarantees that the reviewed literature is comprehensive and pertinent to the subject of flight
price prediction using Generative Artificial Intelligence (GenAI) and deep learning
techniques.

3.1 Search Strategy

A comprehensive search was conducted across multiple reputable databases, including IEEE
Xplore, Google Scholar, Elsevier, ACM Digital Library, Springer Library, and Sci-Hub, to
gather a wide array of literature on the subject. This ensured access to both well-established
and cutting-edge research. The search focused on identifying relevant papers published
between 2020 and 2024, reflecting the most recent advancements in Generative AI, deep
learning, and real-time price forecasting.

Search terms were carefully selected to capture a broad range of topics related to flight price
prediction, while focusing on AI-driven solutions. These terms included combinations such
as “Generative Artificial Intelligence,” “Price Forecasting,” “Deep Learning,” “Neural
Networks,” “Flight Cost Analysis,” and “LSTM (Long Short-Term Memory Networks).”

The search aimed to ensure that studies addressing both foundational AI technologies and
specific applications in flight price prediction were captured. By utilizing these key terms, the
review gathered relevant literature from various disciplines, including machine learning, data
science, forecasting, and aviation economics. Each database was searched using these
combinations, ensuring the inclusion of relevant peer-reviewed articles and research papers
from conferences.

3.2 Inclusion Criteria

After compiling the search results, specific inclusion criteria were applied to select the most
relevant studies for analysis. Research papers were included if they met the following:

- Published in peer-reviewed journals: Ensuring the studies' reliability and credibility.

- Publication Year between 2020 and 2024: Capturing the latest methodologies and
findings due to rapid AI advancements.

- Focus on generative AI and forecasting methods: Emphasizing the use of

Generative AI/other AI-based techniques for improving forecasting in pricing models.

22
 - Written in English: Maintaining consistency and avoiding translation issues.

By adhering to these criteria, the report ensured that the selected literature was both relevant
and of high quality, reflecting the latest advancements in the field.

3.3 Data Extraction and Analysis

Following the identification and filtering of papers using the inclusion criteria, a standardized
data extraction process was employed. Relevant information from the selected studies was
gathered using a structured form designed to capture key aspects of each research work,
ensuring uniformity in the review. The data extraction focused on:

- Flight Rate Prediction Techniques: Key methodologies used for predicting flight
prices, including various models like LSTM, RNN, CNN, and XGBoost, and their
applied AI techniques.

- Generative AI Solutions: Application of Generative AI models such as Generative

Adversarial Networks (GANs) and other deep learning architectures to enhance
prediction accuracy and reliability.

- Gaps in Current Research: Identification of limitations in existing research,

including challenges in real-time forecasting, issues related to data availability or
quality, and model adaptability to market volatility.

The aim of this analysis was to identify common themes, strengths, and limitations across the
studies. By systematically analyzing the data, the report seeks to highlight the current state of
AI-based flight price prediction and identify areas for further research and development. This
process was crucial for understanding the effectiveness of GenAI and deep learning models
in dynamic pricing environments and providing recommendations for future research
directions.

23
CHAPTER 4 - Results

4.1 Overview of the Selected Studies

1. Yuanyuan Guan's research explores the application of Generative AI (GAI) to tackle the
complexities of flight price prediction in a volatile market. Guan demonstrates that leveraging
GAI allows the model to capture intricate patterns in historical airfare data, leading to a more
precise understanding of factors influencing price fluctuations. Integrating deep learning
architectures, such as Recurrent Neural Networks (RNNs) and Long Short-Term Memory
(LSTM) networks, enables the system to process sequential and temporal data, adapting to
changing market trends and enhancing prediction accuracy.

Guan emphasizes the significance of real-time forecasting, achieved through a web-based

platform that connects with live data streams for continuous updates. The use of Generative
Adversarial Networks (GANs) enhances the system's ability to generate realistic future
scenarios based on historical data, providing valuable insights for airlines, travel agents, and
consumers. This dynamic approach allows the model to remain responsive to market trends,
making it an essential tool in the aviation industry.

2. Across the three documents [2-4], we derived several valuable insights directly relevant to
our flight price prediction model:

From Hu Yang et al.'s research on gasoline price forecasting, we learned the importance of
filtering relevant time series data using methods like Lasso regression. The hybrid
Lasso-CNN model they propose screens out irrelevant data, reducing overfitting and
enhancing prediction accuracy. This approach can be adopted to refine our flight price
forecasting model, focusing on impactful variables like seasonality and market trends.

In document [2], Hong et al.'s review of energy forecasting emphasizes leveraging artificial
intelligence and machine learning to identify patterns within historical data. Their focus on
combining probabilistic forecasting and ensemble methods to improve accuracy can be
mirrored in our flight price predictions by integrating different models (e.g., RNN, LSTM)
for more robust outcomes in volatile pricing environments.

Lastly, from Zhang et al.'s paper [4] on electricity price forecasting, we gathered
methodologies for handling price spikes, an issue that could arise with sudden flight price
surges. Their use of frequency analysis and oversampling to predict both normal prices and
price spikes offers a robust framework for improving forecast reliability. These insights
suggest that our model could benefit from applying similar techniques to detect and manage
rare but significant price fluctuations in flight prices.

24
3. Across the two papers [5-6], we gathered key insights to enhance our flight price
prediction model:

Oglakcioglu et al.'s research [5] on optimal flight ticket purchase timing reveals the
effectiveness of learning-to-rank (LTR) methods for predicting the best times to buy tickets.
Their study highlights that the number of days until departure significantly impacts pricing,
varying by route and destination. This approach can optimize our model's predictions based
on specific routes and markets, offering travelers personalized advice on when to purchase
tickets.

Kalampokas et al.'s holistic approach to airfare prediction [6] demonstrates the application of
Machine Learning (ML), Deep Learning (DL), and Quantum Machine Learning (QML)
models. Their analysis of different airlines and destinations shows that diverse AI models
achieve high accuracy (89% to 99%) in predicting airfare. Using multiple AI techniques
addresses the complexities of various routes and airlines. Introducing QML models, a novel
addition to this domain, could further improve prediction accuracy and adaptability.

These insights suggest that integrating LTR, ML, and potentially QML can deliver highly
accurate price predictions while providing users with optimal purchase timing, considering
the dynamic and route-specific nature of airfare pricing.

4. Across documents [7-10], we identified key insights that support our flight price prediction
efforts:

From Supriya Rajankar et al.'s survey [7] on flight pricing using machine learning, we see
that factors like departure time, days left until departure, and time of day significantly impact
ticket prices. They emphasize various models such as Naive Bayes, SVM, and Linear
Regression, noting that SVM achieved the highest accuracy at 80.6%, while other models
also showed promise. This suggests our approach could benefit from testing multiple models
to determine the most accurate predictor of price fluctuations in real-time scenarios.

Heba Mohammed Fadhil et al.'s framework [8] evaluates machine learning techniques for
airfare prediction, highlighting that methods like Decision Trees and K-Nearest Neighbors
(KNN) excel within the Indian airline network, with Decision Trees achieving 89% accuracy.
The recommendation to balance accuracy and efficiency suggests our system should employ
multiple algorithms to enhance prediction accuracy, particularly for frequently changing
flight routes.

In Parwaz Singh Sarao and Pushpendu Samanta’s work [9], flight price prediction using
historical data utilized ML models like Random Forest, Decision Trees, and XGBoost, with
XGBoost yielding the highest accuracy at 84.46%. This supports our decision to use
XGBoost and similar ensemble methods due to their effectiveness in capturing price trends
across various airlines and seasons.

25
Finally, Md. Shaim Hosan Noyon et al.'s thesis [10] demonstrates the effectiveness of
regression-based models like XGBoost and Random Forest for time-series forecasting in
airfare predictions. By focusing on variables such as source, destination, number of stops, and
airlines, they improved accuracy for domestic flights. These findings highlight the potential
of similar ML models in our system, allowing for fine-tuned predictions that dynamically
adjust based on factors like layovers and route complexity.

5. From the final two papers [11-12], we derived key insights applicable to our flight price
prediction model:

Chawla and Kaur's study [11] highlights the advantage of using historical data and data
mining techniques to predict flight price variations. They identified key factors, such as oil
prices, number of competitors, and days until departure, that influence airfare. Their
modeling revealed that flights with fewer competitors experience higher price fluctuations,
while those with more competitors have stable prices. This implies that our model should
account for route competition and other market variables to enhance prediction accuracy.

Bhosale et al.'s research [12] emphasizes the use of machine learning algorithms like Random
Forest, Decision Trees, and Linear Regression to determine the optimal timing for ticket
purchases. Their study showed that certain routes, especially business-heavy ones, exhibit a
non-decreasing price trend as the departure date nears, while other routes show price dips at
specific times. Incorporating these insights into our model can improve recommendations for
ticket purchase timing by focusing on route-specific behaviors and price patterns.

By integrating these findings, our model can further enhance its predictive capabilities by
considering external factors like oil prices and competition while tailoring predictions based
on route-specific trends and competitive behaviors in the airline industry.

4.2 Identified Problems in Implementation

Here are three key challenges and future directions for implementing AI-based prediction
systems in real-time dynamic environments:

1. Data Availability and Quality: Many studies highlight limitations in data availability,
particularly for real-time and complete datasets, such as unsold seats, real-time competition,
or airline-specific algorithms. The absence of critical features can reduce prediction accuracy,
as noted in papers struggling with incomplete or delayed data sources [10-11]. Future AI
models will require more comprehensive real-time data integration from multiple sources,
like airline inventories, competitor behavior, and external economic indicators, to ensure
accurate and timely predictions.

26
2. Model Adaptability to Market Volatility: Several papers indicate the challenge of
adapting AI models to volatile pricing environments, where flight prices are influenced by
numerous, rapidly changing factors like competitor behavior, fuel prices, and market demand.
Even the best models struggle with unexpected market shocks or non-linear events, such as
sudden fuel price hikes or geopolitical events [8, 12]. Future research should focus on
developing adaptive models that can react to these shifts, potentially by integrating
reinforcement learning or hybrid models that can continuously learn from new data.

3. Computational Efficiency and Real-Time Processing: In a dynamic environment, the

challenge lies in making real-time predictions while maintaining high accuracy. Many models
require significant computational resources, and running them in real time may introduce
latency. Algorithms like XGBoost, Random Forest, and Decision Trees, while accurate, are
computationally intensive, complicating real-time implementation [7, 8]. Future AI systems
need to balance computational efficiency with prediction accuracy, possibly by developing
lighter, faster algorithms or improving cloud-based infrastructures that support real-time data
processing.

27
CHAPTER 5 - Discussion

5.1 Synthesis of Findings

Across the reviewed papers, several common themes emerged regarding AI-based flight price
prediction systems:

1. Effectiveness of ML Models: Studies consistently show that machine learning (ML)

models like XGBoost, Random Forest, Decision Trees, and Support Vector Machines
(SVMs) predict flight prices with considerable accuracy. These models effectively learn from
historical data, identify price patterns, and predict based on factors such as departure date,
route competitiveness, oil prices, and the number of stops. Route-specific behaviors are
critical; for instance, business routes tend to follow non-decreasing price trends closer to
departure, while less competitive routes exhibit more variable patterns.

2. Challenges with Data Availability and Quality: A substantial challenge is the

availability and quality of data, which often limits prediction accuracy. Many studies lack
access to real-time data on unsold seats or airline-specific pricing algorithms, resulting in
predictions that may not fully capture dynamic airline pricing strategies. Additionally, AI
models struggle to handle sudden market shifts, such as unexpected spikes in fuel prices or
changes in demand due to external events, indicating a need for more adaptive learning
systems.

3. Real-Time Integration: Integrating these powerful algorithms into real-time dynamic

environments, such as the airline industry, presents both technical and infrastructural
challenges. Most ML models require substantial computational resources. While they
perform well in simulated environments, real-time processing—where flight prices can
change within minutes—demands faster and more efficient systems.

By addressing these themes, AI-based flight price prediction systems can be further refined
and made more effective for real-world applications.

5.2 Implications for Practice

The findings from these studies have profound implications for the travel industry, benefiting
airlines, travel agencies, and consumers alike. For airlines, incorporating AI-based prediction
systems into revenue management strategies allows for more accurate demand forecasting,
optimized seat pricing, and real-time data-driven pricing adjustments. This dynamic pricing
approach maximizes revenue while balancing supply and demand.

For travel agencies, AI-based prediction tools enhance customer satisfaction by providing
accurate fare forecasts and advising on the best times to purchase tickets. This builds
customer trust and loyalty, giving agencies a competitive edge.

28
Consumers also benefit from AI-powered tools, which improve ticket price transparency.
Systems that notify users of optimal purchase times based on historical trends and real-time
conditions help travelers avoid price hikes and save money, making the booking process more
consumer-friendly. These AI models can also be applied to hotel bookings, car rentals, and
other dynamic pricing environments, offering a unified platform for travel cost optimization.

However, the limitations highlighted, such as data availability and the complexity of
integrating AI models into real-time systems, indicate these technologies are not yet fully
mature. Airlines and travel platforms must invest in better data infrastructure, like real-time
tracking of ticket sales, and enhance computational efficiency to fully leverage AI's potential
in real-time price prediction.

5.3 Solutions to the Identified Problems

1. Data Availability and Access:

- Partner with airlines and travel platforms to access real-time data on unsold seats,
customer behavior, and last-minute pricing strategies.
- Establish regulatory frameworks for secure, privacy-compliant data sharing among
competitors to enhance data availability and prediction accuracy.

2. Model Adaptability:

- Integrate reinforcement learning (RL) into AI models to enable continuous learning

and adaptation to new data.
- Develop hybrid AI models that combine deep learning and probabilistic forecasting to
manage both regular trends and unpredictable market shifts.

3. Computational Efficiency:

- Optimize algorithms for faster real-time processing by developing lighter AI models

or utilizing cloud computing infrastructure for large-scale data handling.
- Implement modular AI systems that separate prediction tasks (e.g., handling price
spikes independently) to reduce overall computational demands and accelerate
processing.

4. Data Integration:

- Integrate real-time data from multiple sources such as fuel prices, competitor prices,
and customer demand trends to enhance prediction accuracy.

5. Consumer-Friendly AI Systems:

- Develop transparent AI models that provide users with insights into how predictions
work, enabling consumers to make informed decisions by real-time price forecasts.

29
5.4 Future Research Directions

1. Adaptive Learning Systems: Develop reinforcement learning and self-improving neural

networks that continuously adapt to changing market conditions in real time.

2. Efficiency in Real-Time Processing: Optimize lightweight AI models for real-time

environments and explore quantum computing for faster, complex data processing.

3. Data Privacy and Ethics: Ensure ethical AI use by protecting consumer privacy while
collecting sufficient data for accurate predictions. Design transparent AI systems that allow
users to control how their data is used.

4. Cross-Industry Applications: Apply findings from airfare prediction research to other

dynamic pricing models, such as hotel bookings, ride-sharing, and e-commerce.

5. Real-Time Data Enhancement: Investigate real-time data gathering techniques and

integrate IoT (Internet of Things) devices to improve real-time data collection and prediction
models across sectors.

6. Improving Interpretability: Make AI models more interpretable so stakeholders can

understand and trust predictions, facilitating wider adoption across industries.

30
CHAPTER 6 - Conclusion

This report delves into the development and implementation of AI-based systems for flight price
prediction, emphasizing how Generative AI, machine learning models, and real-time forecasting
techniques address the complexities of dynamic pricing in the airline industry. By synthesizing
findings from 12 key studies, the report highlights the effectiveness of AI systems in accurately
predicting airfare trends through historical data, adaptive learning algorithms, and advanced
techniques like XGBoost, Random Forest, and Reinforcement Learning. Despite these advancements,
challenges like data availability, model adaptability, and computational efficiency need to be tackled
to fully harness AI's potential in real-time environments.

For practical applications, AI systems can greatly benefit airlines, travel agencies, and consumers by
enhancing revenue management, dynamic pricing strategies, and informed ticket purchasing.
Overcoming current limitations, such as integrating more real-time data and developing faster, more
efficient AI models, is crucial for managing the rapid fluctuations in airfare pricing.

Proposed solutions include reinforcement learning, modular AI systems, and partnerships with data
providers to improve prediction accuracy and adaptability. Future research should prioritize adaptive
learning systems, quantum computing, data privacy, and cross-industry applications. Extending
AI-driven price prediction models beyond airlines to sectors like hotels, ride-sharing, and e-commerce
can also offer substantial benefits.

In conclusion, while AI-based flight price prediction systems are effective, their full potential lies in
ongoing development and refinement. By addressing data quality, real-time processing, and model
adaptability challenges, future AI models can provide more precise and dynamic predictions,
transforming price forecasting in rapidly changing markets.

31
CHAPTER 7 - References

[1] Y. Guan, "Flight Price Prediction Web-Based Platform: Leveraging Generative AI for Real-Time
Airfare Forecasting," in Journal of Web Engineering, vol. 23, no. 2, pp. 299-314, March 2024, doi:
10.13052/jwe1540-9589.2325.
keywords: {Costs;Generative AI;Atmospheric modeling;Predictive models;Rendering (computer
graphics);Real-time systems;Stakeholders;Flight price prediction;generative artificial
intelligence;deep learning;real-time forecasting},

[2] H. Yang, X. Tian, X. Jin and H. Wang, "A Novel Hybrid Model for Gasoline Prices Forecasting
Based on Lasso and CNN," in Journal of Social Computing, vol. 3, no. 3, pp. 206-218, September
2022, doi: 10.23919/JSC.2022.0012.
keywords: {Couplings; Social computing; Oils; Time series analysis; Neural networks; Predictive
models; Linear programming; gasoline prices; forecasting; Lasso-CNN; multiple time series},

[3] T. Hong, P. Pinson, Y. Wang, R. Weron, D. Yang and H. Zareipour, "Energy Forecasting: A
Review and Outlook," in IEEE Open Access Journal of Power and Energy, vol. 7, pp. 376-388, 2020,
doi: 10.1109/OAJPE.2020.3029979.
keywords: {Forecasting;Wind forecasting;Load forecasting;Wind power generation;Probabilistic
logic;Industries;Open Access;Energy forecasting;load forecasting;electricity price forecasting;wind
forecasting;solar forecasting},

[4] C. Zhang, Y. Fu and L. Gong, "Short-Term Electricity Price Forecast Using Frequency Analysis
and Price Spikes Oversampling," in IEEE Transactions on Power Systems, vol. 38, no. 5, pp.
4739-4751, Sept. 2023, doi: 10.1109/TPWRS.2022.3218712.
keywords: {Predictive models;Support vector machines;Market research;Artificial neural
networks;Time-frequency analysis;Training;Electricity supply industry;Extended discrete Fourier
transform;price spikes;short-term price forecast;variational mode decomposition},

[5] K. Oglakcioglu, Y. S. Can and F. Alagoz, "Prediction of Optimal Flight Ticket Purchase Timing by
Using Learning To Rank Methods," 2023 Innovations in Intelligent Systems and Applications
Conference (ASYU), Sivas, Turkiye, 2023, pp. 1-6, doi: 10.1109/ASYU58738.2023.10296726.
keywords: {Training;Visualization;Machine learning algorithms;Atmospheric modeling;Urban
areas;Pricing;Predictive models;machine learning;flight ticket price;learning to rank;regression.},

[6] T. Kalampokas, K. Tziridis, N. Kalampokas, A. Nikolaou, E. Vrochidou and G. A. Papakostas, "A

Holistic Approach on Airfare Price Prediction Using Machine Learning Techniques," in IEEE Access,
vol. 11, pp. 46627-46643, 2023, doi: 10.1109/ACCESS.2023.3274669.
keywords: {Airline industry;Atmospheric modeling;Predictive models;Companies;Feature
extraction;Prediction algorithms;Pricing;Airfare price;artificial intelligence;deep learning;machine
learning;prediction model;pricing models;regression;quantum machine learning},

[7] Supriya Rajankar, Neha Sakharkar, 2019, A Survey on Flight Pricing Prediction using Machine
Learning, INTERNATIONAL JOURNAL OF ENGINEERING RESEARCH & TECHNOLOGY
(IJERT) Volume 08, Issue 06 (June 2019),

[8] Heba Mohammed Fadhil; Mohammed Najm Abdullah; Mohammed Issam Younis. "A Framework
for Predicting Airfare Prices Using Machine Learning". IRAQI JOURNAL OF COMPUTERS,

32
COMMUNICATIONS, CONTROL AND SYSTEMS ENGINEERING, 22, 3, 2022, 81-96. doi:
https://doi.org/10.33103/uot.ijccce.22.3.8

[9] Gupta, Shikha & Gupta, Nishi & Publication Hub, Tjcse. (2024). Flight Fare Prediction Using
Machine Learning.

[10] Noyon, Md. Shaim Hosan, Islam, Tanzidul, Islam, Solaiman, Reejon, Md. Refayet Islam, Flight
fare prediction using machine learning, http://hdl.handle.net/10361/23256,

[11] Chawla, Bhavuk, and Ms Chandandeep Kaur. "Airfare analysis and prediction using data mining
and machine learning." Int J Eng Sci Invent (IJESI) 6.11 (2017): 10-17.

[12] Sarao, Parwaz, and Pushpendu Samanta. "Flight Fare Prediction Using Machine Learning."
Available at SSRN 4269263 (2022).

[13] Mahesh, Batta. ”Machine Learning Algorithms-A Review.” International Jour nal of Science and
Research (IJSR).[Internet] 9 (2020): 381-386.

[14] Yazdi, Maryam Farshchian, et al. ”Flight delay prediction based on deep learning and
Levenberg-Marquart algorithm.” Journal of Big Data 7.1 (2020): 1-28.

[15] Lu, W., Li, J., Wang, J. et al. A CNN-BiLSTM-AM method for stock price prediction. Neural
Comput & Applic 33, 4741–4753 (2021). https://doi.org/10.1007/s00521-020-05532-z

[16] Homayoun Ebrahimian, Saeed Barmayoon, Mohsen Mohammadi & Noradin

Ghadimi (2018) The price prediction for the energy market based on a new method, Economic
Research-Ekonomska Istraživanja, 31:1, 313-337, DOI: 10.1080/1331677X.2018.1429291

[17] Hu, Z.; Zhao, Y.; Khushi, M. A Survey of Forex and Stock Price Prediction Using Deep
Learning. Appl. Syst. Innov. 2021, 4, 9. https://doi.org/10.3390/asi4010009

[18] Suiming Guo, Chao Chen, Jingyuan Wang, Yaxiao Liu, Ke Xu, Daqing Zhang, and Dah Ming
Chiu. 2018. A Simple but Quantifiable Approach to Dynamic Price Prediction in Ride-on-demand
Services Leveraging Multi-source Urban Data. Proc. ACM Interact. Mob. Wearable Ubiquitous
Technol. 2, 3, Article 112 (September 2018), 24 pages. https://doi.org/10.1145/3264922

[19] Branda, F.; Marozzo, F.; Talia, D. Ticket Sales Prediction and Dynamic Pricing Strategies in
Public Transport. Big Data Cogn. Comput. 2020, 4, 36. https://doi.org/10.3390/bdcc4040036

[20] Guo, S., Chen, C., Wang, J. et al. Fine-grained Dynamic Price Prediction in Ride-on-demand
Services: Models and Evaluations. Mobile Netw Appl 25, 505–520 (2020).
https://doi.org/10.1007/s11036-019-01308-5

[21] Wei Huang, Jing Hu, Shaorui Zhou, min Zhao, Demand prediction and sharing strategy in
resilient maritime transportation: Considering price and quality competition, Ocean & Coastal
Management, Volume 242, 2023, 106676, ISSN 0964-5691,
https://doi.org/10.1016/j.ocecoaman.2023.106676.

33
[22] Zhang, Kexian, et al. "Forecasting crude oil price using LSTM neural networks." Data Sci.
Financ. Econ 2 (2022): 163-180.

[23] Kim, Hyunsoo, Youngwoo Kwon, and Yeol Choi. "Assessing the impact of public rental housing
on the housing prices in proximity: Based on the regional and local level of price prediction models
using long short-term memory (LSTM)." Sustainability 12.18 (2020): 7520.

[24] M. -Y. Chen, H. -S. Chiang and K. -J. Yang, "Constructing Cooperative Intelligent Transport
Systems for Travel Time Prediction With Deep Learning Approaches," in IEEE Transactions on
Intelligent Transportation Systems, vol. 23, no. 9, pp. 16590-16599, Sept. 2022, doi:
10.1109/TITS.2022.3148269.

[25] Liu, Yanjun, et al. "Understanding urban bus travel time: Statistical analysis and a deep learning
prediction." International Journal of Modern Physics B 37.04 (2023): 2350034.

[26] Wang, Tianyi, et al. "A framework for airfare price prediction: a machine learning approach."
2019 IEEE 20th international conference on information reuse and integration for data science (IRI).
IEEE, 2019.

[27] Ji, Tianbo, et al. "Urban transport emission prediction analysis through machine learning and
deep learning techniques." Transportation Research Part D: Transport and Environment 135 (2024):
104389.

[28] Xiong, Zhi, et al. "Passenger flow prediction of urban rail transit based on deep learning
methods." Smart Cities 2.3 (2019): 371-387.

[29] Amicosante, Andrea, et al. "Predicting costs of local public bus transport services through
machine learning methods." Expert Systems with Applications (2024): 125396.

[30] Patel, Harsh, and Manan Shah. "Energy consumption and price forecasting through data-driven
analysis methods: A review." SN Computer Science 2.4 (2021): 315.

[31] Cheng, Qixiu, et al. "Analysis and forecasting of the day-to-day travel demand variations for
large-scale transportation networks: a deep learning approach." Transportation Analytics Contest,
Tech. Rep (2016).

[32] Cheng, Qixiu, et al. "Analysis and forecasting of the day-to-day travel demand variations for
large-scale transportation networks: a deep learning approach." Transportation Analytics Contest,
Tech. Rep (2016).

[33] Guo, Suiming, et al. "Seeking in Ride-on-Demand Service: A Reinforcement Learning Model
With Dynamic Price Prediction." IEEE Internet of Things Journal (2024).

[34] Singh, Akash, et al. "Prediction of transportation costs using trapezoidal neutrosophic fuzzy
analytic hierarchy process and artificial neural networks." IEEE Access 9 (2021): 103497-103512.

34
[35] Maaouane, Mohamed, et al. "Using neural network modelling for estimation and forecasting of
transport sector energy demand in developing countries." Energy conversion and management 258
(2022): 115556.

[36] Abdou, Hussein A., et al. "A variable impact neural network analysis of dividend policies and
share prices of transportation and related companies." Journal of International Financial Markets,
Institutions and Money 22.4 (2012): 796-813.

[37] Abdou, Hussein A., et al. "A variable impact neural network analysis of dividend policies and
share prices of transportation and related companies." Journal of International Financial Markets,
Institutions and Money 22.4 (2012): 796-813.

[38] Nandal, Mohit, Navdeep Mor, and Hemant Sood. "An overview of use of artificial neural
network in sustainable transport system." Computational Methods and Data Engineering: Proceedings
of ICMDE 2020, Volume 1 (2021): 83-91.

[39] Du, Yu-Ang. "Research on the route pricing optimization model of the car‐free carrier platform
based on the BP neural network algorithm." Complexity 2021.1 (2021): 8204214.

[40] Dougherty, Mark. "A review of neural networks applied to transport." Transportation Research
Part C: Emerging Technologies 3.4 (1995): 247-260.

[41] Heghedus, Cristina, Antorweep Chakravorty, and Chunming Rong. "Neural network frameworks.
comparison on public transportation prediction." 2019 IEEE International Parallel and Distributed
Processing Symposium Workshops (IPDPSW). IEEE, 2019.

[42] Anguita, Juan Gerardo Muros, and Oscar Díaz Olariaga. "Air cargo transport demand forecasting
using ConvLSTM2D, an artificial neural network architecture approach." Case Studies on Transport
Policy 12 (2023): 101009.

[43] Li, Can, et al. "Graph neural network for robust public transit demand prediction." IEEE
Transactions on Intelligent Transportation Systems 23.5 (2020): 4086-4098.

[44] Yao, Shunyu, Yi-Peng Xu, and Ehsan Ramezani. "Optimal long-term prediction of Taiwan’s
transport energy by convolutional neural network and wildebeest herd optimizer." Energy Reports 7
(2021): 218-227.

[45] Baxter, Glenn, and Panarat Srisaeng. "The use of an artificial neural network to predict
Australia’s export air cargo demand." International Journal for Traffic and Transport Engineering 8.1
(2018): 15-30.

[46] Chen, Shu-Chuan, et al. "Improving the forecasting accuracy of air passenger and air cargo
demand: the application of back-propagation neural networks." Transportation Planning and
Technology 35.3 (2012): 373-392.

[47] Ye, Choong-Yeol. "A neural network approach to air cargo fleet assignment." Proceedings of the
KOR-KST Conference. Korean Society of Transportation, 2001.

35
[48] Srisaeng, Panarat, Glenn S. Baxter, and Graham Wild. "Forecasting demand for low cost carriers
in Australia using an artificial neural network approach." Aviation 19.2 (2015): 90-103.

[49] Srisaeng, Panarat, Glenn Baxter, and Graham Wild. "Using an artificial neural network approach
to forecast Australia's domestic passenger air travel demand." World Review of Intermodal
Transportation Research 5.3 (2015): 281-313.

[50] Yu, Bin, et al. "Flight delay prediction for commercial air transport: A deep learning approach."
Transportation Research Part E: Logistics and Transportation Review 125 (2019): 203-221.

[51] Barua, Limon, Bo Zou, and Yan Zhou. "Machine learning for international freight transportation
management: A comprehensive review." Research in Transportation Business & Management 34
(2020): 100453.

36