AR_script.py

# Import the library 
from pmdarima import auto_arima
# Importing required libraries 
import numpy as np 
import pandas as pd 
import matplotlib.pyplot as plt 
import sys
import json
import datetime
  
# Ignore harmless warnings 
import warnings 
warnings.filterwarnings("ignore") 

import os
directory = os.path.join(os.curdir, "sine_data")
seq_len = 200
dims = [1, 2, 4, 6, 9]

for dim in dims:
    file = f'sine_123_1000_{dim}.csv'
    data = np.loadtxt(os.path.join(directory, file), delimiter = ",",skiprows = 1)
    # dim = file.split('_')[3][0]
    print(f'dim is {dim}')

    data = pd.DataFrame(data)

    if dim == '1':
        data = data[1]
        best_model = auto_arima(data, start_p = 1, start_q = 1, 
                            max_p = 3, max_q = 3, 
                            start_P = 0, seasonal = True, 
                            d = None, D = 1, trace = True, 
                            error_action ='ignore',   # we don't want to know if an order does not work 
                            suppress_warnings = True,  # we don't want convergence warnings 
                            stepwise = True)           # set to stepwise
        forecast = pd.DataFrame(best_model.predict(n_periods = seq_len))
        gen_list = forecast.values.tolist()
        x = datetime.datetime.now()

        timestamp = x.strftime("%d%m%y_%Hh%M")
        filepath = f'synthetic_sines/arima_sine_{dim}_{seq_len}_{timestamp}.json'
        with open(filepath, 'w') as file:
            json.dump(gen_list, file)

    else:
        # Fit auto_arima function to dataset
        generated_sines = []
        for feature in data.columns:
            feat = data[feature]
            best_model = auto_arima(feat, start_p = 1, start_q = 1, 
                                    max_p = 3, max_q = 3, 
                                    start_P = 0, seasonal = True, 
                                    d = None, D = 1, trace = True, 
                                    error_action ='ignore',   # we don't want to know if an order does not work 
                                    suppress_warnings = True,  # we don't want convergence warnings 
                                    stepwise = True)           # set to stepwise
            forecast = pd.DataFrame(best_model.predict(n_periods = seq_len))
            generated_sines.append(forecast.values.tolist())
            # Save generated data to csv
        x = datetime.datetime.now()

        timestamp = x.strftime("%d%m%y_%Hh%M")
        filepath = f'synthetic_sines/arima_sine_{dim}_{seq_len}_{timestamp}.json'
        with open(filepath, 'w') as file:
            json.dump(generated_sines, file)