Prediction Method Using Neural Network

​

In order to do time series prediction of crypto-currency value using neural network, the library used here is keras. First the time series value is converted to data frame for data analysis using  as.data.frame(matrix()) functions. The data is divided into two equal parts for training and testing. The training and testing labels are then generated using to_categorial() function. The model used here is keras_model_sequential(). The model is layers are made using the softmax activation. The model is compiled using the 'mse' loss parameter, the optimizer used is adam and the metric is accuracy. The model is trained using the training data and training labels and the the prediction is done on test using predict() function. Then the prediction is scaled from probabilistic value to real value using the average crypto-currency price.

 

                              Loss                   Accuracy     

                         Bitcoin             0.0011                  75.00                        

  Dogecoin        0.001                  98.46

Ethereum        0.0022                 50.00

Litecoin        0.001                  99.67

Steller        0.0023                  50.00

BNB        0.0175                  50.00

Polkadot        0.0195                  26.67

Solana        0.0165                  32.65