Machine-Learning-Stock-Analysis

Use sentiment analysis of past articles and analysis of technical indicators to predict short term trends of stocks. Target specific industries and companies following a max market cap.

With about 600 data points for each industry, we are at an average of around 66% accuracy.

Work in Progress

• Implementing MACD analysis
• Collecting more data
• Implementing Java Rest API for GUI

APIs Used

• Alpha Vantage
• Stocknews

Libraries Used

• Numpy pip-install

pip install numpy

• Sklearn install

pip install -U scikit-learn

• os
• os.path
• datetime
• requests
• json
• time

config.py

Your config file should include the following:

api_token = '**** YOUR STOCK NEWS API ****'
alpha_token = '**** YOUR ALPHA VANTAGE API ****'

Usage

Once you have your data in a file called 'data', all you have to do is use processor.py to create your features and targets, then use learner.py to fit and train a learner for classification. Example can be seen in the "Testing" section in the learner.py file.

Data Form

In order to use these files, you must collect your own data (news and any historical/technical data on stocks). The news collected from the StockNews API were in the form of JSON files as follows:

{
    "data": [
        {
            "news_url": "example.com",
            "image_url": "example.com/image.jpg",
            "title": "example title",
            "text": "example text",
            "source_name": "source",
            "date": "Fri, 12 Jun 2020 09:05:00 -0400",
            "topics": [
                "PressRelease"
            ],
            "sentiment": "Neutral",
            "type": "Article",
            "tickers": [
                "TEST"
            ]
        },
        {
            "news_url": "example.com",
            "image_url": "example.com/image.jpg",
            "title": "example title",
            "text": "example text",
            "source_name": "source",
            "date": "Fri, 12 Jun 2020 09:05:00 -0400",
            "topics": [
                "PressRelease"
            ],
            "sentiment": "Neutral",
            "type": "Article",
            "tickers": [
                "TEST"
            ]
        }
    ]
}

and the data on stocks collected from Alpha Vantage in the form of JSON files as follows:

{
    "Meta Data": {
        "1: Symbol": "APEX",
        "2: Indicator": "Moving Average Convergence/Divergence (MACD)",
        "3: Last Refreshed": "2020-06-22",
        "4: Interval": "daily",
        "5.1: Fast Period": 12,
        "5.2: Slow Period": 26,
        "5.3: Signal Period": 9,
        "6: Series Type": "open",
        "7: Time Zone": "US/Eastern"
    },
    "Technical Analysis: MACD": {
        "2020-06-22": {
            "MACD_Signal": "0.0465",
            "MACD": "0.0526",
            "MACD_Hist": "0.0061"
        }
    }
}

License

MIT