Chris Pollett > Students > Kabra
Print View
[Bio]
[Blog]
[CS297 Proposal]
[Deliverable1]
[Deliverable2]
[Deliverable3]
[Deliverable4]
[CS 297 Report-PDF]
[CS298 Proposal]
[CS 298 Report-PDF]
[CS 298 Presentation-PDF]
|
CS298 Proposal
A Neural Net For Stock Trend Predictor
Sonal Kabra (sonal.kabra@sjsu.edu)
Advisor: Dr. Chris Pollett
Committee Members: Robert Chun, Paul Thienprasit
Abstract:
The Neural Net Stock Trend Predictor is an approach to developing a stock trend predictor. Stock prediction using computers is also known as algorithmic trading. Algorithmic trading is the process of using a computer program to follow a defined set of instructions for placing a trades at speeds, complexity, or frequency beyond what a human trader can do. Algorithmic trading can be based on timing, price, quantity or any mathematical model. For this project I will use Qandl stock data sets. I will build multi-layer neural networks to predict buy, sell or hold classification for a single or a portfolio of stocks for time scales of one day.
CS297 Results
- Measured return, volatility of stock by inputing stock ticker
- Implemented the simple neural net for estimating starting month of give dataset
- Implemented the machine learning K- nearest neighbor and Decision tree algorithm for stock prediction
- Implemented the above machine learning algorithms to predict for the portfolio of stock
Proposed Schedule
Week 1: February 08, 2017 - February 13, 2017 | A program to calculate simple moving average, Exponential moving average, Relative strength index |
Week 2: February 14, 2017 - February 20, 2017 | Deliverable 1 due |
Week 3, 4: February 21, 2017 - March 6, 2017 | A program to calculate Earning per share growth, revenue growth |
Week 5: March 7, 2017 - March 13, 2017 | Deliverable 2 due |
Week 6, 7: March 14, 2017 - March 27, 2017 | A Convolutional Neural Net for predicting price for a single stock using above features |
Week 8: March 28, 2017 - April 3, 2017 | Deliverable 3 due |
Week 9, 10: April 4, 2017 - April 17 , 2017 | Extend the Convolutional Neural Net to predict for the portfolio of stock |
Week 11: April 18, 2017 - April 24, 2017 | Deliverable 4 due |
Week 12, 13: April 25, 2017 - May 8, 2017 | Deliverable 5 - CS298 Report |
Week 14: May 9, 2017 - May 15, 2017 | Deliverable 6 - Presentation |
Key Deliverables:
- Software
- A program to calculate simple moving average, exponential moving average, relative strength index on stock data obtained from Quandl website
- A program to calculate earning per share growth, revenue growth on economic data obtained from Quandl website
- A Convolutional Neural Net for predicting price for a single stock using above features
- Extend the above Convolutional Neural Net to predict for the portfolio of stock
- Report
- CS298 Report
- CS298 Presentation
Innovations and Challenges
- A multi layer Convolutional Neural Net works better with the 2 dimensional or more data. Implementing it for the one dimensional stock data is challenging as the data is sparse.
- There are very few papers on how neural net can be implemented for prediction of the portfolio of a stock. So doing research on this and implementing it is a challenge.
References:
Barry Johnson, Algorithmic Trading and DMA: An introduction to direct access trading strategies, 4Myeloma Press (17 February 2010)
Luca Di Persio, Oleksandr Honchar, Artificial Neural Networks Approach to the Forecast of Stock Market Price Movements. International Journal of Economics and Management Systems, 1, 158-162, (2016)
Perry Kaufman. Trading Systems and Methods, Wiley; 5 edition, (2013) |