Stock Price Indicator

This is a python script for predicting stock's adjusted close price using data from Quandl. This script will take your current stock open price, highest price, and volume to make a prediction for the adjusted close price of the day.

Important

• Authentication: Replace quandl.ApiConfig.api_key = 'YOURAPIKEY' to your own Quandl API key in data.py. Strongly recommend replacing this with your own API key since the current API key in data.py is shared with the public.
• Libraries: Must have quandl, yahoo-finance, parse, numpy, pandas, and matplotlib installed to your python interpreter.
• Installation: The installation instructions are for OS X and Linux users. Windows users need to install a terminal that runs Unix commands.
• Usage: Using this script during nonmarket hours will give you a prediction with data leakage because you are predicting a known adjusted close price. However, you can use manual inputs option to test or estimate using your own open, high, and volume data during nonmarket hours.
• Known Bug: Since there are limited queries for real-time stock information from Yahoo finance, sometimes it will query the data from previous trade day or even return Null. Therefore, you may want to put down your real time stock information manually(using Manual inputs) for accurate predictions.
• Algorithm: The final model is trained by Bagging with Decision Trees as the base estimator. However, you can use the Random Forest algorithm which did not create a significant improvement for most stock.
• For more details of this project, please refer to report.pdf.

Installation

1. Download and unzip this package.
2. Replace authentication key to your own Quandl API Key at line 3 in data.py.
3. Install quandl, yahoo-finance, parse, numpy, pandas, and matplotlib to your python. You can use pip install [library] in your terminal to install.

How to use this script##

1. Modify input.txt with your stocks' ticker symbols and leave everything else as default for recommendations or put down your own customized data (Please read Definition for input.txt for more details).
2. Open your terminal. Run python main.py in this directory. It will take about a minute to predict the adjusted close price.
3. Use the result to justify your buying and selling.

Definition for input.txt##

Main Inputs

• Ticker symbols: Put down all your ticker symbols and separate each of them with ,. (e.g. GOOG,FB).
• Recommendation: Put down yes for recommendations to buy or sell for each trained stock. (Default: no)
• Manual inputs: Put down yes to enable Stock data inputs. (Default: no)
• Stock data: Put down the stock data that you want to predict manually with (ticker symbols, open price, current highest price, volume) format and separate each of them with ;. For example, (goog,741.86,742.0,2980700);(fb,126.89,128.80,15691100)). Be aware that open price and current highest price need to have 2 decimals and volume has to be a whole number. Remember to put down yes at Manual inputs to enable this feature.
• Data size: Put down the number of data you want to train. The number must be an integer. (Default: 600)
• Random Forest : Put down yes to enable Random Forest algorithm. (Default: no)

Developer Inputs

• Developer Mode: Set it on to enable all the developer inputs. (Default: off)
• Show best_estimator: Set it on to print out the best estimator and it's parameters. (Default: off)
• Show estimated graph: Set it on to plot a graph that compares the predicted data and actual data. (Default: off)
• Start_date: Put down data training starting date with yyyy-mm-dd format or default(data from the past three months).
• End_date: Put down data training ending date with yyyy-mm-dd format or default(data from the past three months).
• Prediction by date: Set it on to enable Predict_date inputs and disable today's prediction and recommendation. (Default: off)
• Predict_date: Put down the date you want to predict with yyyy-mm-dd format and separate each of them with ,. For example, 2015-09-21, 2016-09-15. Important: Predict dates that do not exist in the quandl database will create an error. Remember to set Prediction by date to on in order to enable this feature.
• Data preprocessing: Set it off to disable data preprocessing. (Default: on)

Debugging Inputs

Please ignore all the debugging inputs since it's for debugging.