This repository contains my data analytics project for a Kaggle hackathon, where I focused on stock price prediction for the 'Close' and 'Strategy' columns using the 'Open' and 'Volume' features.
Leaderboard Ranking: I'm proud to share that I achieved the 32nd position out of 274 participants on the Kaggle leaderboard for this hackathon.
In this hackathon, I aimed to predict stock prices and develop a strategy (Hold, Buy, or Sell) based on historical data. I used the following techniques and models:
- Linear Regression: I used linear regression as my final model for predicting the 'Close' price of stocks. Although I experimented with ARIMA initially, it didn't perform as well as linear regression. You can find the ARIMA code in the repository as well.
- Ensemble Learning: To predict the 'Strategy' (Hold, Buy, or Sell), I leveraged ensemble learning techniques.
- XGBoost: I employed the XGBoost model as one of the ensemble base models.
- Soft Voting: The final prediction for the strategy was made using soft voting among the ensemble models.
To run the code and reproduce the results, follow these steps:
- Clone this repository to your local machine.
- Navigate and open the Kaggle notebook for step-by-step analysis.
- Review the code and comments to understand the data preprocessing and modeling process.
- Execute the notebooks to generate predictions for the 'Close' and 'Strategy' columns.
Happy coding!