Python AI experiments for Bitcoin price prediction

import numpy as np
import pandas as pd 
import matplotlib.pyplot as plt
from datetime import datetime
from autots import AutoTS

data = pd.read_csv('BitcoinHistoricalData.csv')
print("Shape of Dataset is: ", data.shape, "\n")
print(data.head())

# Convert DataFrame column type from string to datetime
data['Date'] = pd.to_datetime(data['Date'])

# Sort by date column
data = data.sort_values('Date').reset_index(drop=True)

# Select the column "Price" for daily price
# Price strings have commas as thousands separators so you will have to remove them
#   before the call to float
data['Price'] = (data['Price'].str.split()).apply(lambda x: float(x[0].replace(',', '')))
data['Price'] = data['Price'].astype(float)

# Soften data w/ a moving average on price
movingAvgWindow = 30
data['Price'] = data['Price'].rolling(window=movingAvgWindow).mean()
data = data[movingAvgWindow:]


model = AutoTS(forecast_length=120, frequency='infer', ensemble='simple', drop_data_older_than_periods=3000)
model = model.fit(data, date_col='Date', value_col='Price', id_col=None)
prediction = model.predict()
forecast = prediction.forecast

# Save, in case you need to store results
# forecast.to_csv('forecast.csv', index=True)

# Load stored forecast, in case you need to load previous results
# forecast = pd.read_csv('forecast.csv')

# Add column headers
forecast.columns = ['Date', 'Price']

# Convert DataFrame column type from string to datetime
forecast['Date'] = pd.to_datetime(forecast['Date'])

print("Bitcoin Price Prediction")
print(forecast)

# Draw it
plt.figure(figsize=(24,10))
plt.plot(data['Date'].values,     data['Price'].values,     label = 'Real Bitcoin Price',      color = 'red')
plt.plot(forecast['Date'].values, forecast['Price'].values, label = 'Predicted Bitcoin Price', color = 'blue')
plt.xlabel('Date')
plt.ylabel('Price ($)')
plt.legend()
plt.show()