STOCK PRICE Indicator

Capstone project for Udacity Data Science Nanodegree 2021

This project provides a stock price indicator which downloads historical stock price information from the Alpha-Vantage platform. In order to download data from Alpha-Vantage please refer to https://www.alphavantage.co/support/#api-key. The API key has to be set in process_data.py for the parameter A_key. This version uses my personal API key which will be deactivated after review of the project.

Once the data has been downloaded the application performs a technical analysis calculating a bunch of indicators like the simple moving average, Bollinger bands, momentum, daily returns and cumulative returns. Based on these information the project further provides a machine learning pipeline to predict future stock prices.

Different machine learning algorithms have been tested first. The application of the different models can be found in the Jupyter notebook Stock_price_indicator.ipynb. Once everything has been analyzed and tested in the Jupyter Notebook the final application has been implemented in python scripts and a Webapp. For a detailed description of the project please refer to Project_documentation_20210531.pdf

There are two ways to access the stock price analysis and prediction:

1. Via a flask web app web_app.py which launches a web server on http://127.0.0.1:5000/
2. Using the scripts ETL.py and ML.py
   • ETL.py provides an interface to enter a time period and a list of stock ticker symbols in order to download and analyse the historical data
   • ML.py provides an interface to enter a time period and a list of stock ticker symbols for which future stock prices shall be predicted. The ticker symbols must be a subset of the previously selected stock data

Installation

Clone this repo to the preferred directory on your computer using git clone https://github.com/jochenruland/stock_price_indicator. Then start either the Webapp /app/web_app.py or run the two scripts /app/ETL.py and /app/ML.py .

You must have installed the following libraries to run the code:

sys pandas numpy matplotlib alpha_vantage.timeseries datetime sklearn flask werkzeug json plotly sqlite3

Program and dataset files:

Main files

• app/data_wrangling/process_data.py: module that includes the class StockDataAnalysis and some helper functions
• app/data_wrangling/train_classifier.py: module that includes the class ModelStockPrice
• app/ETL.py: ETL pipeline which extracts the data from Yahoo finance, calculates and joins features and safes the result to indicators.db
• app/ML.py: Machine learning pipeline which loads the stock data from indicators.db, instantiates a model object which is then fitted and used for prediction
• app/web_app.py: Starts the Python server for the Webapp.

File structure

app
|- web_app.py # Flask file that runs app
|- ETL.py
|- ML.py
|- indicators.db # Database saving results from ETL.py
|- AAPL.csv # Sample data download for Apple
|- BABA.csv # Sample data download for Alibaba
|- GOLD.csv # Sample data download for Barrick Gold
|- GOOG.csv # Sample data download for Alphabet
|- SPY.csv  # Sample data download for S&P 500 Market Index
|- templates
| |- index.html # main page of web app
| |- base.html # formatting page of web app
| |- post.html # result page of web app
|- data_wrangling
| |- process_data.py
| |- train_classifier.py
README.md
Project_documentation_20210531
LICENSE.txt
Stock_price_indicator.ipynb