K-Means Clustering on Stocks

This project involves the manual implementation of the K-Means clustering algorithm applied to stocks. The primary objective of this project was to conduct personal research and gain a deeper understanding of clustering techniques and their applications in financial data analysis.

Try it out here!

Project Overview

The K-Means algorithm is a popular unsupervised learning method used for clustering data into distinct groups based on their features. In this project, I manually implemented the K-Means algorithm from scratch, without relying on pre-built libraries, to better understand its inner workings and nuances.

Features

1. Interactive Stock Selection:

   • Users can select multiple stocks from the NIFTY 50 index and add custom stocks for analysis
   • Flexible timeframe selection to analyze different time periods
   • Ability to specify the number of clusters (k) for the analysis

2. Real-time Clustering:

   • Dynamic clustering of selected stocks based on user parameters
   • Interactive visualization showing cluster assignments and centroids

3. Data Analysis Tools:

   • Detailed view of financial metrics for each stock
   • Cluster-wise analysis showing common characteristics
   • Visual tracking of clustering iterations and convergence

Implementation Details

1. Data Collection:

   • Historical stock price data was collected for NIFTY 50 stocks using the yfinance library.
   • The data includes daily closing prices, which were used to calculate various financial metrics.

2. Data Preprocessing:

   • The collected data was cleaned and preprocessed to handle missing values and normalize the features.
   • From the data, we derived the following features: Mean Returns, Volatility, and Sharpe Ratio.

3. Manual K-Means Implementation:

   • Initialize the centroids randomly from the data points.
   • The algorithm iteratively assigned each data point to the nearest centroid and updated the centroids based on the mean of the assigned points.
   • The process was repeated until the centroids stabilized or a maximum number of iterations was reached.

How to Run

1. Clone the repository.
2. Install the required packages using pip install -r requirements.txt.
3. Run the Jupyter notebook KMeans.ipynb to see the implementation and results.
4. Launch the Streamlit application using the command streamlit run KMeans.py to interact with the clustering results.