Miscellaneous Machine learning Projects

• Tech Stack: Image classification, Regression, Time series analysis/Prediction, Resgression
• Github URL: Project Link

I worked on 4 different projects focusing on different areas within Machine learning. Experimentation and fine tuning of different machine learning models are carried out for each project.

• Project 1:

Machine Learning Models for a Network Intrusion Detector

The objective of this project is to design a network intrusion detector by using 4 different machine learning models(SVM,logistic regression, kNN,FCNN) comparing their performance.This is a classification problem to be solved by taking network data of known connections and determining whether they are safe connections or attacks.

• Project 2:

Stock Prediction Models Using Deep Learning

The goal of this project is to create a time-based stock price prediction tool using deep learning models(FCNN,LSTM and CNN). This system should be able to predict the price of a stock based on the past 7 days. As additional features this project can also predict stock prices for a continuous future time period(Ex: for 5 consecutive days). Best model was chosen among the 3 models and tested on several different datasets.

• Project 3:

House Price Estimation Using a Combined Deep Learning Model

The goal of this project is to create a home price prediction system using a composite model to use both numeric data and images, by with functional API. The composite model consists of a CNN(for images) and a Fully Connected Neural Network(for numeric data).

• Project 4:

Solving (Wild) Tic-Tac-Toe using Minimax Search

The goal of this project is to implement Minimax for both Tic-Tac-Toe and Wild Tic-Tac-Toe to demonstrate the use of adversarial search. The developed Minimax algorithm was used to test the results under each settings (i.e., optimal vs optimal; optimal vs random). Where optimal players make correct game decisions every-time. Random players are prone to make mistakes. All such observations can be used to validate the working efficiency of the Minimax algorithm