Stock Predictor API

LEGAL DISCLAIMER

This project is for education purpose only and the available API's is for personal use only. The data collected using yfinance is under Yahoo!'s terms and conditions.

Introduction

Personal project that enables users to get stock prediction data for the next 30 days. The project is based on Meta's Prophet project. The project uses upto 5 year's worth of opening price data for a stock and predicts the opening price for the next 30 days. A frontend website is developed to visualize the data, Link, Repo.

Project Support Features

REST API end point returns last 6 months opening price and projected opening price for the next 30 days.

Installation Guide

• Clone the repository
• The master branch is the most stable branch at any given time, ensure you're working from it.
• Create a virtual environment using python -m venv name_of_your_env
• Run pip install -r requirements.txt all dependencies
  • Windows user might also need to add pywin32==306 to the requirements.txt file.
• Alternatively you can access the API which is hosted on render, here.

Project Structure

• app.py contains the endpoints and handle's the requests.
• forecast.py machine learning script to predict the price.
• requirements.txt contains the packages used for this project.

Deployment

The API is hosted on render.com. The free tier is used for the deployment which will cause the application to shutdown after 1 minute of inactivity. For the first request, it can take upto 1 minute for the application to start up again.

URL: https://stock-prediction-flask.onrender.com

REST API Endpoints

The REST API endpoints are described below. The body and the response are in JSON format.

Homepage

Use this endpoint to ping the server to start the application. The application will timeout after 1 minute of inactivity.

Request

GET /

curl https://stock-prediction-flask.onrender.com/

Response

{"result":"connected"}

Predict

Request

POST /predict

Body

The endpoint accepts a json body with the value of the ticker key to be the stock ticker. For example, 'aapl' for Apple Inc.

Example:

{
"ticker": "aapl" 
}

curl --header "Content-Type: application/json" \
  --request POST \
  --data '{ "ticker": "aapl" }' \
  https://stock-prediction-flask.onrender.com/predict

Response

For a valid ticker value, the response will contain a result key which holds the response array. The response will contain up to past 6 months of opening price, under date and past, forecast will be null for this period. After that data, the forecast data will be under date and forecast, past will be null for this period.

{ "result": 
        [{
		"date": "Mon, 31 Jul 2023 00:00:00 GMT",
	        "forecast": null,
	        "past": 195.537995145902
	},{
		"date": "Tue, 01 Aug 2023 00:00:00 GMT",
	        "forecast": null,
	        "past": 195.7175151124968
	}, ....., 
	{ 
		"date": "Sat, 03 Feb 2024 00:00:00 GMT",
	        "forecast": 199.13563380619226,
	        "past": null
	    }, {
	        "date": "Sun, 04 Feb 2024 00:00:00 GMT",
	        "forecast": 199.48080692540407,
	        "past": null
	    }, ....
	  ]
}

For an invalid ticker, the server will return an response error code 406. This could be due to invalid ticker value or recent IPO's which might not be available.

{
	"errorType":  406,
	"message":  "No data"
}

Technologies Used

Python 3.11.1 was used to develop this application.

Package	Version
flask	3.11.1
yfinance	0.2.35
prophet	1.1.5
pandas	2.2.0

Limitations

The machine learning model, prophet, only utilizes on piece of data which is the stock's historical opening price. A stock price depends on multiple factors and an acurate prediction using this API is not possible. This project should not used as a factor to invest in a potential stock and is for education purpose only.