Time-Series-Prediction-with-TensorFlow

A series of different Time Series modelings experiments with various models to predict Bitcoin price

Table of Content

• The Problem
• Goal
• Project Main Steps
• Data Visualization
• Modeling
• Prediction
• Conclusion
• Software and Libraries

The Problem

Python Bitcoin is widely used cryptocurrency for digital market. It is decentralised that means it is not own by government or any other company.Transactions are simple and easy as it doesn’t belong to any country.Records data are stored in Blockchain.Bitcoin price is variable and it is widely used so it is important to predict the price of it for making any investment.

Goal

This project focuses on the accurate prediction of cryptocurrencies price using tensorflow.

Project Main Steps:

• Get time series data (the historical price of Bitcoin)
• Format data for a time series problem
  • Creating training and test sets
  • Visualizing time series data
  • Turning time series data into a supervised learning problem (windowing)
  • Preparing univariate and multivariate (more than one variable) data
• Evaluating a time series forecasting model
• Setting up a series of deep learning modelling experiments
  • Dense (fully-connected) networks
  • Sequence models (LSTM and 1D CNN)
  • Ensembling (combining multiple models together)
  • Multivariate models
  • Replicating the N-BEATS algorithm using TensorFlow layer subclassing
• Creating a modelling checkpoint to save the best performing model during training
• Making predictions (forecasts) with a time series model
• Creating prediction intervals for time series model forecasts

Data Visualization

Price of Bitcoin from 4 November 2014 to 2 Feb 2022	Price of Bitcoin from 4 November 2014 to 2 Feb 2022 Train&Test

Modeling

Data -> Format data for a time series problem -> build a model -> Evaluating -> Making predictions (forecasts)

Model	mae	mse	rmse	mape	mase
naive_model	1127.009888	2704483.50	1644.531494	2.823276	0.998678
model_1_dense_w7_h1	1140.201172	2764825.75	1662.776489	2.847246	1.008707
model_2_dense_w30_h1	1233.389771	3079271.75	1754.785522	3.085911	1.084139
model_3_dense_w30_h7	2419.106689	11996798.00	2770.479492	6.062185	2.125433
model_4_CONV1D	1144.036255	2792920.00	1671.203125	2.851182	1.012100
model_5_LSTM	1187.783813	2923671.50	1709.874756	2.962411	1.050802
model_6_multivariate	1139.666504	2739723.25	1655.210938	2.848582	1.008234
model_8_NBEATs	1180.154175	2880262.50	1697.133789	2.962500	1.044052
model_9_ensemble	1157.829346	2824341.75	1680.577759	2.891772	1.024302

Comparing the Performance by mae of Each of Our Models

Prediction

Prediction of bitcoin price with interval values	Price of Bitcoin Prediction to Future

Conclusion

In this capstone project, I took Bitcoin prices from CoinDesk, analyzed them, applied various models to fit the data, and forecasted the models. The predictions we have made here are not financial advice. Furthermore, by now, we should be well aware of just how poor machine learning models can be at forecasting values in an open system - anyone promising you a model which can "beat the market" is likely trying to scam you, oblivious to their errors or very lucky.

Software and Libraries

This project uses the following software and Python libraries: