Stock and Cryptocurrency Price Prediction usi...

Stock and Cryptocurrency Price Prediction using ARIMA Model

Author Name : Pradeep Kumar K., Sankhadeep Bakshi, P Naveen Prabhath, Rohith V, Amrutha G Krishna

 

DOI: https://doi.org/10.56025/IJARESM.2022.107980

 

ABSTRACT

There has been a huge improvement in investing in general and many people have started learning and investing in different types of investments. Stocks and cryptocurrency have seen the most growth of all investment methods, this is due to their high profitability and economic growth. But, stocks are prone to market crashes based on real-world events and issues, and crypto is extremely volatile, due to its variable nature and uncertain usage. Investing is a risky business, but as technology has improved many new tools have been developed to evaluate these risks, one of these is using machine learning and artificial intelligence technology. These techniques allow the investors to reduce the risk involved in trading and investing. Over the past few years, developers have been able to create highly accurate prediction algorithms.

In our project, we used the "ARIMA model" for price prediction. Due to its unique ability to handle time-series data and its ability to extract information from a large data-set more accurately, we chose ARIMA. We used the Yahoo Finance API to extract all the historical data. Yahoo Finance has current and historical data for nearly all the companies and cryptocurrencies. All historical data were extracted using the API and divided into two major categories: crypto and stocks. Then we pre-processed the data and fed it to our ARIMA model for prediction. In our model, we implemented a user-defined function to simulate actual market trading, i.e., buy when the market is low and sell when the market is high.

Therefore, our model uses the historical data and performs multiple simulations to identify patterns. We also included a parameter in our model for multiple outputs with slight thresholds, this was to ensure that the prediction would always be accurate as the market is volatile. Our model's accuracy after the forecast was 88-94%. We then stored the data in a local database. Then we developed the website and connected the back-end to these results, so that when users use the site, they can easily access these results and invest their money with lower risks.  

Keywords: ARIMA, Cryptocurrencies, Machine Learning, Prediction, Shares, Volatile