daradecic

DATASET_NAME = 'DogsVsCats'
SEGMENTS = {'training_set': True, 'test_set': False}

def DogsVsCats(path: str) -> Dataset:
    root_path = os.path.abspath(os.path.expanduser(path))
    dataset = Dataset(DATASET_NAME)
    dataset.load_catalog(os.path.join(os.path.dirname(os.path.dirname(path)), 'catalog.json'))
    
    for segment_name, is_labeled in SEGMENTS.items():
        segment = dataset.create_segment(segment_name)

daradecic

# Max lag order
max_p = 10
# To store RMSE
errors = {}

for p in range(1, max_p + 1):
    # Train and predict
    model = AR(df_train['y']).fit(maxlag=p, dynamic=False)
    preds = model.predict(
        start=len(df_train),

daradecic

def train_and_plot(maxlag):
    model = AR(df_train['y']).fit(maxlag=maxlag, method='mle')
    forecasts = model.predict(
        start=len(df_train),
        end=len(df_train) + len(df_test) - 1,
        dynamic=False
    )
    
    parameters = model.params.to_dict()
    for k, v in parameters.items():

daradecic

# Train/test split
df_train = df[:-10]
df_test = df[-10:]

# Plot
plt.title('Random dataset train and test sets', size=20)
plt.plot(df_train['y'], label='Training data')
plt.plot(df_test['y'], color='gray', label='Testing data')
plt.legend();

daradecic

import numpy as np
import pandas as pd
from sklearn.metrics import mean_squared_error
from statsmodels.tsa.ar_model import AR

import matplotlib.pyplot as plt
from matplotlib import rcParams
from cycler import cycler

rcParams['figure.figsize'] = 18, 5

daradecic

plt.title(f'Airline passengers Triple Exponential Smoothing predictions\nRMSE (mul, add) = {np.round(rmse_mul_add, 2)} | RMSE (mul, mul) = {np.round(rmse_mul_mul, 2)}', size=20)
plt.plot(df_test['Passengers'], color='gray', label='Testing data')
plt.plot(predictions_mul_add, color='orange', label='Predictions Multiplicative trend, Additive seasonality')
plt.plot(predictions_mul_mul, color='red', label='Predictions Multiplicative trend, Multiplicative seasonality')
plt.legend();

daradecic

# Multiplicative trend / Additive seasonality model
model_mul_add = ExponentialSmoothing(df_train['Passengers'], trend='mul', seasonal='add', seasonal_periods=12)
results_mul_add = model_mul_add.fit()
predictions_mul_add = results_mul_add.forecast(steps=24)

# Multiplicative trend / Multiplicative seasonality model
model_mul_mul = ExponentialSmoothing(df_train['Passengers'], trend='mul', seasonal='mul', seasonal_periods=12)
results_mul_mul = model_mul_mul.fit()
predictions_mul_mul = results_mul_mul.forecast(steps=24)

daradecic

from statsmodels.tsa.holtwinters import ExponentialSmoothing

# Additive model
model_add = ExponentialSmoothing(df_train['Passengers'], trend='add')
results_add = model_add.fit()
predictions_add = results_add.forecast(steps=24)

# Multiplicative model
model_mul = ExponentialSmoothing(df_train['Passengers'], trend='mul')
results_mul = model_mul.fit()

daradecic

from statsmodels.tsa.holtwinters import SimpleExpSmoothing

# Model parameters
span = 12 
alpha = 2 / (span + 1)

# Train the model
model = SimpleExpSmoothing(df_train['Passengers'])
results = model.fit(smoothing_level=alpha, optimized=False)
predictions = results.forecast(steps=24)

daradecic

# Train/test split
df_train = df[:-24]
df_test = df[-24:]

# Plot
plt.title('Airline passengers train and test sets', size=20)
plt.plot(df_train['Passengers'], label='Training data')
plt.plot(df_test['Passengers'], color='gray', label='Testing data')
plt.legend();