Before we move anywhere close to decentralised machine learning (or federated learning) lets quickly understand what centralisation and decentralisation is. As the name suggests centralised refers to one body which manages or controls the system, like a bank: which manages and controls your money, and only they (the banks and their top executives) have the … Continue reading Decentralised Machine Learning: Federated Learning

This is part 3 of end-to-end blog series on DS project. Please follow the first and second blog for better understanding this step. Once the model has been developed its time to serve it to end users to see how its performing on the live data. Flask framework would be used here to serve the … Continue reading Part 3 End-to-End DS project: Deployment

This blog is the part 2 of series end-to-end DS project. Follow the first blog in case you havent. Lets convert the df_final to dataset and labels so that we can use it in the score_classifier function (we'd using it for scoring different models) dataset = df_final.iloc[:,:-1] #X labels = df_final.iloc[:, -1] #Y print(f'shape of X: {dataset.shape}') … Continue reading Part 2 End-to-End DS project: Modelling

This is a 3 part blog where we are going to look at the end-to-end Data Science project. In this part we'll be highly focusing on EDA process, followed by modelling process and then in the end we'll deploy the model using Flask framework. We'll also look at the deployment using Heroku PaaS (Platform as … Continue reading Part 1 End-to-End DS project: EDA

For this blog, I'll be using Fb's API (called Graph API) to fetch the data, and then dump it into MongoDB. I've divided this task in two parts:  1. Scraping the data from Fb  2. Dumping it into the MongoDB  Scraping: Scraping Fb is not an easy task, and it has tightened its anti-bot system … Continue reading Data extraction using an API and storing into MongoDB

In this blog we'll be looking at the hybrid model of CNN + LSTM. The objective here would be to build an intuition of its working rather than jumping into the code directly. But at the same, I also encourage to go through code, you can start with simple vanilla architecture then work your way … Continue reading Hybrid model (CNN + LSTM) for detecting aggressive content

In this blog we'll see the basic data pipeline for NLP based project. The reason we need to have a neatly defined pipeline is so that the entire process could be further divided into different entities which works independently of each other, and if we have any future issues or any features to add then … Continue reading Data pipeline for an NLP based project