train.py

import os
import pandas as pd
import itertools
from helper import *

import warnings
warnings.filterwarnings('ignore')

def main():
    config_plot()
    cwd = os.getcwd()
    train = pd.read_csv(cwd+'\\results\\data\\training_dataset.csv',index_col='date')   # 89 ---> 15
    print(train.shape)
    valid = pd.read_csv(cwd+'\\results\\data\\validation_dataset.csv',index_col='date') # 14 ----> 19 
    print(valid.shape)
    if not os.path.exists(os.path.join(cwd+'\\results', "models")):# check if the directory exists
      os.mkdir(cwd + "\\results\\models") # create the directory if it doesn't exist
   #  else :
   #    #remove all files in the directory
   #    for filename in os.listdir(cwd + "\\results\\models"):
   #      file_path = os.path.join(cwd + "\\results\\models", filename)
   #      try:
   #          if os.path.isfile(file_path) or os.path.islink(file_path):
   #              os.unlink(file_path)
   #      except Exception as e:
   #          print(f'Error deleting {file_path}: {e}')

   #  models  = ['NN','MPL','simpl_RNN','LSTM','BI_LSTM','CNN' ]
    models  = ['simpl_RNN','LSTM','BI_LSTM','CNN' ]
    
    for model in models: 
         hyp_param = get_optimal_hyperparamaters(cwd+'\\results\\tuning_summary.csv', model)
         print(hyp_param[2])
         if model == 'NN':
            model_ = NN_construction(layer_units=hyp_param[4],input_shape=hyp_param[2])
         elif model == 'MPL':
            model_ = MLP_construction(layer_units=hyp_param[4],layer_num=hyp_param[3],input_shape=hyp_param[2])    
         elif model == 'simpl_RNN':
            model_ = simpl_RNN_construction(layer_units=hyp_param[4],layer_num=hyp_param[3],input_shape=hyp_param[2])           
         elif model == 'LSTM':
            model_ = LSTM_construction(layer_units=hyp_param[4],layer_num=hyp_param[3],input_shape=hyp_param[2])           
         elif model == 'BI_LSTM': 
            model_ = BI_LSTM_construction(layer_units=hyp_param[4],layer_num=hyp_param[3],input_shape=hyp_param[2])          
         elif model == 'CNN': 
            model_ = CNN_construction(layer_units=hyp_param[4],layer_num=hyp_param[3],input_shape=hyp_param[2])         
         # elif model == 'CNN_LSTM': 
         #    model_ = CNN_LSTM_construction(layer_units=hyp_param[4],layer_num=hyp_param[3],input_shape=hyp_param[2])
         else : 
            raise ValueError('Invalid model name: {}'.format(model))
         
         train_dataset = windowed_dataset(series=train.values,window_size=hyp_param[2])
         train_date    = pd.to_datetime(train.index)[hyp_param[2]-1:]
         # print(train_date.shape)
         train_actual  =  train[hyp_param[2]-1:]
         # print(train_actual.shape)
         valid_date    = pd.to_datetime(valid.index)[24:]
         # print(valid_date.shape)
         valid_actual  =  valid[24:]
         # print(valid_actual.shape)
         
         if hyp_param[2]   == 12 : 
            valid_dataset = windowed_dataset(series=valid[12-1:].values,window_size = hyp_param[2])
         elif hyp_param[2] == 18 :
            valid_dataset = windowed_dataset(series=valid[6-1:].values,window_size = hyp_param[2])
         else:
            valid_dataset = windowed_dataset(series=valid[:].values,window_size= hyp_param[2])
         
         history = model_train(model = model_ ,dataset=train_dataset,validation_data=valid_dataset,learning_rate=hyp_param[5],patience=5)
         plot_training_validation_loss(model=model ,history=history) 
         train_predicts = model_forecast(model=model_,series=train.values,window_size =hyp_param[2] )
         print(train_predicts.shape)
         if hyp_param[2]   == 12 :
            valid_predicts = model_forecast(model=model_,series=valid[12+1:].values,window_size =hyp_param[2])
            print(train_predicts.shape)
         elif hyp_param[2] == 18 :
            valid_predicts = model_forecast(model=model_,series=valid[6+1:].values,window_size =hyp_param[2])
            print(valid_predicts.shape)
         else:
            valid_predicts = model_forecast(model=model_,series=valid[1:].values,window_size =hyp_param[2])
            print(valid_predicts.shape)
         
         

         rmse = loss_comp(valid_predicts, valid_actual.values.reshape(-1),loss = 'rmse')
         mae = loss_comp(valid_predicts, valid_actual.values.reshape(-1),loss = 'mae')
         
         print(f'------------------------{model}--------------------')
         
         print(f'The bast paramaters are : input shape = {hyp_param[2]}, numberof layers = {hyp_param[3]},number of units = {hyp_param[4]}, optimal learning rate  = {hyp_param[5]}')
         print(f'Validation loss (RMSE) of {model} model is:{rmse}')
         print(f'Validation loss (RMSE) of {model} model is:{mae}')
         
         print('----------------------------------------------------')
         
         models_saver(model=model_ ,name = model)
         plot_series(train_date,(train_actual, train_predicts),fig_name= f'{model} model_fit',legend=["Actual_values", "Fited_values"])#,title=f" Fited vs Actual values of {model} model (training dataset)")
         plot_series(valid_date,(valid_actual, valid_predicts),fig_name= f'{model} model_forcast_validation',legend=["Actual_values", "Forecasted_values"])#),title=f" Fited vs predicted values of {model} model (validation dataset)")
      
if __name__ == "__main__":
    main()