DatevDuration.txt

from sklearn.preprocessing import PolynomialFeatures

# Shuffle the data
indices = np.random.permutation(len(Full_Solar_Data))
shuffled_days = np.array(days)[indices]
shuffled_durations = np.array(Full_Solar_Data['duration.s'])[indices]

# Reshape the shuffled_days array
shuffled_days = shuffled_days.reshape(-1, 1)

# Create polynomial and sinusoidal features
poly = PolynomialFeatures(degree=4)  # You can adjust the degree as needed
X_poly = poly.fit_transform(shuffled_days)
X_with_sin = np.concatenate((X_poly, np.sin(shuffled_days)), axis=1)

# Fit linear regression model
model = LinearRegression()
model.fit(X_with_sin, shuffled_durations)

# Predict durations
predicted_durations = model.predict(X_with_sin)

# Sort the shuffled_days for better visualization
sorted_indices = np.argsort(shuffled_days[:,0])
sorted_days = shuffled_days[sorted_indices]
sorted_predicted_durations = predicted_durations[sorted_indices]

# Plot the shuffled data and regression line
plt.scatter(shuffled_days, shuffled_durations, color='skyblue', edgecolor='black')
plt.plot(sorted_days, sorted_predicted_durations, color='red', linewidth=2)

plt.title('Date vs Duration (with Polynomial and Sinusoidal Features)')
plt.xlabel('Date')
plt.ylabel('Duration')
plt.show()