Solid State Technology

Volume: 63 Issue: 6
Publication Year: 2020

Financial Supervision And Management System

Using Ml Algorithm
Atishay Jain#1, Yash Sharma*2, Dr. Kaushal Kishor#3
#1
Under Graduate Student ABES Institute of Technology, Ghaziabad, U.P, India
*2
Under Graduate Student ABES Institute of Technology, Ghaziabad, U.P,
India
#3
Associate Professor in IT Department, ABES Institute of Technology, Ghaziabad, U.P, India
ABSTRACT

In this paper, we aim to put forward highly beneficial features in banking and money tracking web-based
application. We make use of machine learning algorithms mathematical tools available for Python which is our
programming language of choice. As per the ML part of our paper, we have used Random Forest Algorithm as a
base on which we have built a proposed algorithm which will predict the user’s financial portfolio. For the
prediction to take place we are taking the help of tools like RNN, LSTM, and MLP. We are aiming to also include a
few modern technologies in order to make the financial supervision system, a more usable and interactive
experience i.e., by including features like Virtual assistant and a chatbot using technologies like NLTK. This paper
shows a proposed algorithm which can be taken into consideration when designing algorithms for banking and
financial needs.

Keyword- Random Forest Algorithm, NLTK, RNN, LSTM, MLP.

I. INTRODUCTION
In most countries like India, which is a developing country, money management, accounting, budget
execution or other financial things are still being in an incredibly old way. We currently use so much
powerful stuff as compared to the previous generation such as a smartphone, laptops etc but still, we
use the traditional ways of managing our finances. As we have seen and done a lot of research on
how small businesses work in our country, we have concluded that people still lack technological
advancements when it comes to knowing how to predict their business’s future trend or market. They
lack prediction tools which are not easily accessible in order to see future trends. This has had
deleterious effects on the functioning of their public expenditure management, they also tend to
spend on other areas which might not be that essential for their business currently without analysing
on what areas they should focus and put their money on.

The lack of authentic and disciplined revenue and expenditure data for budget planning, monitoring /
controlling, and expenditure control and reporting has negatively affected the budget management of
these small businesses. India is a developing nation with a lot of young potentials when it comes to
manufacturing great products for the entire world, but they lack a good budget and resource
management. We look forward to taking a step and help these small businesses for doing so by
providing them with great tools for the prediction of market trends and their own business trends.

18974
 Solid State Technology
Volume: 63 Issue: 6
Publication Year: 2020

II. FEATURE EXTRACTION

A. Deep Learning: Deep Leaning, the most important methodology, is the solution of all the
complex inputs.
1. Prediction of Financial Market: Using the historical data and different parameters of the
current market scenario, the neural networks in deep learning predicts the financial values. As
deep learning takes too much data into consideration, taking the hidden layers as well, the
accuracy and efficiency of the prediction improve up too many extents [17]. Some techniques
used in prediction like:
a. Recurrent Neural Network (RNN): RNN is used for organizing data in a sequential order
e.g., time-series database [16].
b. Long Short-Term Memory Models (LSTM): In order to help with longer memory so that
the forecasted time horizon can be longer, LSTM technique is used.
c. Multilayer Perceptron (MLP): MLP has consisted three layer Input layer, Hidden layer
and Output layer. This is also suitable for time series forecasting because it is: Non-linear
modelling, Multi-step forecasting etc.
2. Security of Finance: As taking care of the security regarding finance, deep learning is the best
approach. Deep learnings are used in such a way to automate the searching data process for
anomalies that could be a security threat.

B. Machine learning and Algorithm based Machine Reasoning -

Machine Learning is a subset of data science that uses statistical insights and makes predictions. The
main motive of this methodology is to provide the stabilised, efficient and accurate prediction [11].
In terms of getting the efficiency and accuracy, the algorithm named Random Forest, which is a big
win has been used up, supervised learning which has an advantage of solving classification and
regression relating problems which is helping a lot in making the almost perfect prediction. Random
Forest algorithm builds multiple decision trees and combines them to give a stable and efficient
prediction. Random forest algorithm reduces the overfitting as if there are so many trees present in
the forest, the classifier will not be able to overly fit the model. Random Forest is simple, fast to
execute, flexible in terms of stable prediction. Random Forest algorithm uses Sklearn (formerly
known as Scikit- learns), which is a free machine learning python library which is a biggest helping
hand in terms of making a stable and efficient prediction.
Sklearn some built-in functions for random forest algorithm: -
 n_estimators- stabilizes the prediction
 max_features- splits the nodes
 min_sample_leaf - determines the minimum number of leaves required to split an internal node.

18975
 Solid State Technology
Volume: 63 Issue: 6
Publication Year: 2020

The Mathematics involved in Random Forest:

1. Regression Problems: When to solve regression problems, using the Random Forest
Algorithm [9].

In the formula, MSE stands for Mean Square Error. The formula has a range from i=1 to N.
This formula calculates the distance of each node from the predicted actual value, helping to
decide which the better decision for your forest is. Here, yi is the data point value we are
testing at a certain node and fi is the value given by the decision tree.
2. Classifications Problems: When applying Random Forests based on classification data, we
should keep this thing in mind that we are often using the Gini index, or the formula applied
to decide how nodes on a decision tree branch [10].

Gini Index here calculates the amount of probability of a specific feature. It’s value ranges
between 0 and 1.This formula uses the class and probability to determine the Gini value of
each branch on a node, determining which of the branches is more likely to occur. Here, pi
represents the class relative frequency we are observing in the dataset and c represents the
number of classes.

C. NLTK: Natural Language Toolkit is a toolkit for developing Python programs to work with
human language data. It provides interfaces such as Word Net, along with text processing
libraries suite for a different piece of works like classification, tokenization, etc. for NLP
(Natural Language Processing) libraries. NLTK is used in the development of virtual assistants
and chatbots wherein the virtual assistant, the responses were recorded in accordance with the
casual users relating problem. Casual users facing problem relating with access to the system and
its usage can help to sort it out very easily [15].

III. LITERATURE REVIEW

Following are a few literature reviews papers we discovered and read. These papers contain similar
implementations and ideas whereas they have some or the other limitations.
We reviewed the following papers in order to come to a few conclusions as well as define their
limitations when compared to our approach:
Javier A. Varela, Nobert When (2017). In this paper, the author comes up with various analysis of
how risk can be managed using PyOpenCL in Python. The author emphasis on mathematical
optimizations in managing risk and how we can make use Jupyter Notebook in creating an
environment for analysing the state of the financial portfolio [2].

18976
 Solid State Technology
Volume: 63 Issue: 6
Publication Year: 2020

Peng Zhang, Yuxiang Gao, Xiang Shi (2018). In this paper, we read that is the Quant Cloud, trade
and design algorithmic studies were signified using Python. They took consideration of Machine
Learning algorithms and used the cloud for data exchange facility [3].

Puneet Mathur (2019). In this paper, the author takes into consideration various other fields in which
Python has shown wonderful success using ML technologies, these fields include healthcare
industry, Oil & Gas, government, transportation etc. with keeping that in mind, he comes to various
conclusions about how python can be especially useful when it comes to financing [4].

Anish Gupta (2019). In this paper the author emphasis upon how data can be classified in various
structured, semi structured and unstructured forms. He also gives examples of companies like Yahoo
and Google who finance through supervised and unsupervised machine learning algorithms. Also, in
this paper they analysed and predicted stocks using ML algorithms [5].

G. Biau,” Analysis of a Random Forests Model. In this paper we concluded that how Random Forest
works, the author describes it in a very elaborative terms through a model. The author describes how
randomization in it takes place, how sparsity happens etc [10].
We concluded a few limitations from the existing model:

Lack of Prediction feature which is a core part of our paper. These given papers do not mention how
we can use Machine Learning Algorithms such as Logical Regression or Random Forest in order to
predict the customer’s future financial status.

1. Lack of modern requirements for a business-oriented web app such as the implementation of
chatbot and a virtual assistant which will enable customers to interact with the system in a more
intrusive way.
2. No special mentions of graphical representation of the user’s financial data by using technologies
such as XlsxWriter.

IV. PROPOSED SYSTEM OVERVIEW

The given model shows how the proposed system will work, all the steps we will take to implement
it and they use them to predict the data based on the performance of it.

18977
 Solid State Technology
Volume: 63 Issue: 6
Publication Year: 2020

Figure 1: Proposed System overview

V. IMPLEMENTATION

For the implementation part of there, we are using the Random Forest Machine Learning Algorithm
as for the base on which we are going to propose our algorithm.

PROPOSED ALGORITHM
Basic steps to proposing the Algorithm:

Step 1- Descriptive Analysis such as identifying inputs by the user, identifying categorical &
numerical values, eliminating missing values, etc.
Step 2- Using the Random Forest techniques; create a benchmark solution which models the data in
an efficient manner.
Step 3- Performance estimation by dividing your train data set into train and validate sections (70:30
as a ratio).
Step 4- Importing necessary libraries, tools and train the data set. Libraries and tools such as Pandas,
NumPy, Sklearn, random etc.
Step 5- Create summary datasets and in columns classify them accordingly.
Step 6- Identify different variables according to the proposed dataset model such as ID, Target
variables, categorical variables or numerical variables.
Step 7- Treating the variables with missing values and create a flag for them. Mark them as false or
indicate them with a coloured coded tag.
Step 8- Impute missing values and pass them into the modelling process by using the random forest
to predict the classes of them.
Step 9- Check the performance of it and make predictions accordingly.

Let us consider a scenario of US revenue data report.

18978
 Solid State Technology
Volume: 63 Issue: 6
Publication Year: 2020

This table is consisting of different dataset values which are as follows to estimate future values of
US revenues:

Revenues- The inputs taken in this table are in either increasing or decreasing order from one of the
US revenues reports.
3-mo MA (month Moving Average)- These inputs are the averages of 3 months, calculated by the
inbuilt math module. e.g. (=Average (Jan: Mar)).
5-mo MA (month Moving Average)- These inputs are the averages of 5 months, calculated by the
inbuilt math module. e.g. (=Average (Jan: May)).

Here, in the table below we can see some blocks left in the starting of the 3-mo MA and 5-mo MA,
these gaps represent the average of 3 month and 5-month revenues, respectively.

For example, the average of Jan, Feb, March is represented in the month of March in the 3-mo MA
column in the below table, similarly for the 5-mo MA.

($5.0(Jan) + $8.0(Feb) + $7.0(mar))/3 = $6.7, which is represented in 3-mo MA.

Similarly, ($5.0(Jan) +$8.0(Feb) + $7.0(Mar)+$8.0(Apr)+$8.0(May))/5=$7.2,which is represented

in 5-mo MA.

TABLE 1: The US revenue & 3 and 5 mo Moving Averages The graph shows the output of the
input data set
Revenues 3-mo MAs 5-mo MAs
Jan $5.0
Feb $8.0
Mar $7.0 $6.7
Apr $8.0 $7.7
May $8.0 $7.7 $7.2
Jun $9.0 $8.3 $8.0
Jul $7.0 $8.0 $7.8
Aug $9.0 $8.3 $8.2
Sept $5.0 $7.0 $7.6
Oct $7.0 $7.0 $7.4
Nov $5.0 $5.7 $6.6
Dec $8.0 $6.7 $6.8

Figure: 2 Moving Average Prediction

Where Revenue, 3 -MO MAs, 5- MO MAs

The graph represents the input variables, in which we input revenues starting from the month of
January to December.

18979
 Solid State Technology
Volume: 63 Issue: 6
Publication Year: 2020

Here we have predicted the graph using our proposed algorithm in which these are the following
points to note down:
1. X-axis is representing the months (Jan to Dec).
2. Y-axis is representing the capital in US $(Dollars).
3. The 3-month MA varies to a greater degree, with a prominent increase or decrease in historic
revenues in comparison to the 5- month MA.
4. The 5-month MA varies to a greater extent in comparison to 3-month MA.

The above model is a clear representation of the working of our proposed algorithm based on the
random forest algorithm. In this, the revenue is considered as a possible feature for the tree in the
random forest which results in 3 month and 5-month predictions. Revenue also represents a part of
the historical data of US financial portfolio.

VI. ADVANTAGES

1. This paper would be able to predict future trends for businesses and customers.
2. The paper would be able to diagrammatically represent the given data into the form of excel
oriented charts and diagrams.
3. The virtual assistant and the chatbot would help the customers navigate through the application
and easily do the task they want to do.

VII. APPLICATION
The Financial Management System is utilitarian in many different applications. Especially in
collecting information, budgeting, reporting, predicting, analysing etc.in so many sectors. This can
be run and used in any kind of financial sectors. Lots of people use these types of application but not
able to be used efficiently as they failed to communicate and work rightly. That is why this system is
proposed to go beyond in terms of helping every kind of person whether they are casual or not. AI
virtual assistants and chatbots can help in communicating with the users so that they can work over
their problems easily and efficiently.
Secondly, no system has been proposed till, yet which can predict the financial market whether it
will rise or fall. But using this system this prediction will take place.

VIII. CONCLUSION
Financial supervision system all studies the casualties taken by the casual user in accordance to
control it efficiently and easily through AI virtual assistant and chatbots. It also studies the prediction
to be predicted through the old and trending markets and predicts whether the financial market will
seek to high or low in the upcoming time. Through this review paper, we try to communicate all the
problems which are usually faced by the people whether they are literate or illiterate.

REFERENCES
[1] “Yuxing Yan”, “Using Python for Finance 101”. Published in January 2013, at Canisius College.
SSRN Electronic Journal.

18980
 Solid State Technology
Volume: 63 Issue: 6
Publication Year: 2020

[2] “Javier A. Varela, Nobert When”, “Real-Time Financial Risk Measurement of Dynamic
Portfolios with Python and PyOpenCL”. DOI: 10.1145/3149869.3149872, 7th Workshop, November
2017.

[3] “Peng Zhang, Yuxiang Gao, Xiang Shi”, “Quant Cloud using Python”, Published in July 2018 at
2018 IEEE International Conference on Software Quality, Reliability and security (QRS), Lisbon,
Portugal. The project name is: Big Data Infrastructure for Quantitative Finance on the Cloud.

[4] “Puneet Mathur”, “ML application Using Python”, DOI: 10.1007/978-1-4842-3787-8_9, January
2019. [Book retrieved from Research gate]. ISBN: 978-1-4842-37-86-1.
[5] “Anish Gupta, Manish Gupta, Prateek Chaturvedi”, “Investing Data with ML using Python”, July
2019, at International Conference on Quality, Reliability, Infocom Technology and business
Operation, New Delhi.

[6] “Yuan Tao”, “Analysis on Financial Fraud cases by Python”, DOI: 10.2991/icesed-19.2020.8, at
Proceedings of the 2019 International Conference on Education Science and economic development
(ICESED 2019). Published in Advances in Economics, Business and Management Research, volume
116.
[7] X.H. Hu. Financial Management, Shanghai: Shanghai Lixin Accounting Press, 2009.
[8] L. Zhang Financial Management System Mode, Research Management, 2006.
[9] L. Breiman. Random forests, Machine Learning, 45:5–32, 2001.

[10] G. Biau ,” Analysis of a Random Forests Model”, Journal of Machine Learning Research 13
(2012) 1063-1095.

[11] Tyagi D., Sharma D., Singh R., Kishor K., “Real Time ‘Driver Drowsiness’ & Monitoring &
Detection Techniques”, International journal of Innovative Technology And Exploring Engineering,
Vol. 9, Issue 8, ISSN 2278-3075, pp 280-284, June 2020

[12] “Divya Yadav, Dr. Thillairasu”, “Stock Market Analysis Using Python” in IJIRT. Paper ID:
150009, Publication Volume & Issue: Volume 7, Issue 2, Page(s): 147 – 150.

[13] “David Varga”, “Fintech, the new era of financial services”, for the project: “The Triple-
Bottom-Line impact of Fintech”. This paper was published on 03 November 2017.

[14] “Kranthi Sai Reddy Vanukuru”, “Stock Market Prediction Using Machine Learning”, published
during November 2018 at Sreenidhi Institute of Science & Technology.

[15] E. Loper, S. Bird, “Workshop on Effective tools and methodologies for teaching natural
language processing and computational linguistics”, ETMTNLP '02: Proceedings of the ACL-02 -
Volume 1, Pages 63–70, July 2002

[16] Sherstinsky, Alex, “ Fundamentals of Recurrent Neural Network (RNN) and Long Short-Term
Memory (LSTM) network”, Physical D: Nonlinear Phenomena, Vol 4, March 2020.

18981
 Solid State Technology
Volume: 63 Issue: 6
Publication Year: 2020

[17] M. Melvin,W. Panb, P. Wikstrom, "Retaining alpha: The effect of trade size and rebalancing
frequency on FX strategy returns", Journal of Financial Markets, Volume 51, November 2020.

18982