Introduction: The Best Stock Visualizing and Predicting Site

Stock is trend for investment and financial management. People can make a great amount of earnings from it. Simultaneously they can also loose a great amount of money from it. Wherefore, a great stock visualize tool can boost trading efficiency. Moreover, it would be best if there's an add-on predicting tool to assist investors. This post details how to build a stock visualizing and predicting site with twelvedata and streamlit.

Contents

• How to Start (build process)
• Details of Approach
• Results
• References

How to Start (build process)

To create and activate a conda environment (Python 3.7.10):

conda create -n stock python==3.7.10

To check if your environment is successfully created:

conda env list

Activate conda environment:

conda activate stock

To install the project's dependencies:

pip install -r requirements.txt

After creating your environment, visit twelvedata to obtain your API key:

To check if your API key is valid, run the example below.

from twelvedata import TDClient
# Initialize client - apikey parameter is requiered
td = TDClient(apikey="YOUR_API_KEY_HERE")

# Construct the necessary time serie
ts = td.time_series(
    symbol="TSLA",
    interval="1min",
    outputsize=10
)

# Returns pandas.DataFrame
print(ts.as_pandas())

You are all set to build The Best stock visualizing and predicting site.

Details of Approach

This project is divided into two parts, the workflow is display as below:

1. Generating LSTM models

   In this part we generate LSTM for different stocks, by using tensorflow and keras.

   • Obtain the data with twelvedata API
   • Scale the data with MinMaxScaler with the range 0 to 1
   • Re-organize the data, we choose to use 60 previous days of data to predict the 'close' price today
   • Feed the data for training the LSTM model
   • Finally, we output the model for later on usage

2. Build main site

   The main site is built with different sections of information, including historical data, predicted price etc.

Results

• The front page of the site

• In the left pane, users can choose the stock with a certain time duration they want to visualize and predict

• In the left column, we display 1 year of historical data

• In the top right column, we display realtime data of selected stock. Users can also select the time interval they like

• In the bottom right column, the predicted stock price is displayed. Users can choose the number of days of predictions they want to visualize

References

• API
  • twelvedata
  • twelvedata github
• Libraries: Tensorflow, sklearn, streamlit, plotly
• Data used：APPL historic stock data, AMZN historic stock data, TSLA historic stock data