Cryptocurrency-Prediction

Inspiration

The concept of Cryptocurrency has captured the imagination of all young minds across the globe. Cryptocurrency is a digital currency in which transactions are verified and records maintained by a decentralized system using cryptography, rather than by a centralized authority. Thus this currency does not recognize any man-made national boundary. Its value is not decided by the GDP of any one country. It is not controlled by any country, as such represent equality in true spirit and thus is truly global. The value of the cryptocurrency keeps fluctuating because of supply and demand, investor and user sentiments, government regulations, and media hype. Many citizens look down upon cryptocurrency as a tool for wealth generation because of these unprecedented fluctuations in the value of the cryptocurrency with time. Thus the importance of early and accurate gauging of value fluctuation of the cryptocurrency and its prediction is obvious and huge for the financial well-being of the investors and users. Hence we took up to build a program to predict the value of the cryptocurrency with scientific tools, a set of preceding values of the currency, and other data available. We are developing the model with the least error in the predicted value and the actual value on a given date and time.

How We Built

We built it using a machine-learning algorithm named LSTM(Long-Short Term Memory). We used Tensor-Flow API for model Training and Prediction. Flask has also been used to connect the model built and trained to the web application to be used by the user. And finally, used HTML and CSS to create the front-end Web Application.

Chanllenges We Faced

Though the creation of the model was easy, due to fluctuating prizes, creating the model with minimum loss was a challenging part. We had to try a different number of layers and optimizers to achieve was we have.

Our Future

Here, We have predicted the prices based on only the past few days' prices. But the price depends upon many other factors like the GDP of the country and so on. So we would try and train our model on various other factors that affect the cryptocurrency's carrying prices.