To get started, install the required packages:
pip install -r requirements.txt
Start the application
python main.py
''' Enter a stock ticker: A000390 Similar tickers to A000390: [ (1, 'A000850', '2016-06-21', '2016-11-01', 0.9998597495307318), (2, 'A001460', '2017-09-04', '2018-01-19', 0.9998423548323742), (3, 'A001720', '2016-06-22', '2016-11-02', 0.9998410764022848), (4, 'A001130', '2017-08-31', '2018-01-17', 0.9998346156897776), (5, 'A001530', '2016-06-14', '2016-10-25', 0.9998210524366685) ] '''
- The CLI interface allows users to input a stock ticker and receive results.
data_loader.pyis used to fetch data from the existingsample_stock_prices.csv.- Initially, cosine similarity is used for calculating similarity. It observes the correlation between two vectors, normalized for computation.
- For each stock, data from 2015-01-02 to 2019-12-30 is segmented into 90-day periods, and similarity is calculated for all cases.
- Currently, a brute force approach is used for these calculations. As the number of cases and stocks increase, the time to compute and produce results will also increase. There's a need to explore and test different methods for more efficient processing.
- Presently, calculations are performed for all cases, but sampling over intervals can be an alternative approach.
- The implementation is class-based with OOP, enhancing maintainability and ease of adding new features. It involves:
- Loading data
- Attaching a model
- Receiving input stock values and generating output