Skip to content
No description, website, or topics provided.
Jupyter Notebook TeX Other
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
.ipynb_checkpoints
screenshots
AMD.csv
GOOGL.csv
README.md
Stokive.key
Stokive.pdf
echarts.min.js
index.html
jquery.min.js
model.pkl
my_model.h5
notebook.tex
output_6_1.png
predicted.json
stockive.ipynb
stockive.py
stockive2.ipynb
true_data.json
webserver.py

README.md

Project name: Stockive

Team members:

The problem chosen:

Stock prediction system

Context:

The record of stock price is a time series, which reflects the performance of a corporate. Although there are a lot of complicated theoretical financial features to evaluate the stock price, it is still difficult to predict the price technically. However, as the deep learning technology is getting advanced, the opportunity to gain a steady fortune from the stock market is amplifying. And we want to design a good algorithm which is based on deep learning method to predict the stock price and help the investors to gain steady profits. Moreover, we aim to apply our algorithm into a practical system that can be used in real investment environment, which is of great importance to help in maximizing the profit of stock option purchase while keeping the risk low.

Approach:

We use stock price data from AMD and GOOGLE to train the model in order to achieve a more comprehensive algorithm. To achieve this, we choose fifty days as a window size and one day as a predict size. The goal is to integrate the information of fifty days' stock prices to predict the next day's. From the original dataset, we split 20% as testing dataset and 80% as training dataset. The algorithm we use here is Recurrent Neural Networks(RNN). More specifically, it is Long Short Term Memory networks(LSTMs), a special kind of RNN, capable of learning long-term dependencies, from which we can capture the long time trend of the stock price.

Data: Historical AMD and GOOGLE Stock Price since 2009, Real time stock price from Alpha Vantage APIs

  • We firstly applied these two datasets as the train and test data to modify the model. According to the performance, by adjusting the parameters and feature selection, a better model will be delivered based on test data and everyday prediction.
  1. AMD.csv
  2. GOOGLE.csv
Field Data Type Description
Date date time The date of the stock price
Open float The price when market just opened
High float The highest price in the day
Low float The lowest price in the day
Close float The price when market is closed
Volume int The trading volume of the market

Source files link: https://www.kaggle.com/gunhee/amdgoogle

  • To follow up, we aim to apply real time stock data into the system using Alpha Vantage APIs which provide realtime and historical global equity data in 4 different temporal resolutions: (1) daily, (2) weekly, (3) monthly, and (4) intraday. Daily, weekly, and monthly time series contain 20+ years of historical data. By applying the API, it can return a JSON file containing the real time stock information. Based on the four resolutions, we are able to retrieve the data details as requirements.

eg. (preview of intraday data)

{
"Meta Data": {
"1. Information": "Intraday (5min) open, high, low, close prices and volume",
"2. Symbol": "MSFT",
"3. Last Refreshed": "2019-02-11 10:40:00",
"4. Interval": "5min",
"5. Output Size": "Compact",
"6. Time Zone": "US/Eastern"
},
"Time Series (5min)": {
"2019-02-11 10:40:00": {
"1. open": "105.7400",
"2. high": "105.8200",
"3. low": "105.7100",
"4. close": "105.7300",
"5. volume": "161340"
},
"2019-02-11 10:35:00": {
"1. open": "105.5900",
"2. high": "105.7400",
"3. low": "105.4600",
"4. close": "105.7350",
"5. volume": "192407"
}

} }

System screenshot

- -

You can’t perform that action at this time.