build_first_forecasts.py

import pandas as pd
from kats import models

# Selection your base model
def build_model(model, kts):
    if model == "prophet":
        return models.prophet.ProphetModel(kts, params=models.prophet.ProphetParams())
    elif model == "theta":
        return models.theta.ThetaModel(kts, params=models.theta.ThetaParams())
    elif model == "holtwinters":
        return models.holtwinters.HoltWintersModel(kts, params=models.holtwinters.HoltWintersParams())
    elif model == "linear":
        return models.linear_model.LinearModel(kts, params=models.linear_model.LinearModelParams())
    elif model == "quadratic":
        return models.quadratic_model.QuadraticModel(kts, params=models.quadratic_model.QuadraticModelParams())
    print(f"{model} is not a right input , using default prophet")
    return prophet.ProphetModel(kts, params = models.prophet.ProphetParams())

# Load the data
dfp_data = pd.read_csv(folder + "kaggle_playground_january_2022/train.csv")
dfp_data["category"] = dfp_data.apply(lambda row: row["country"] + "-" + row["store"] + "-" + row["product"], axis=1)
categories = dfp_data["category"].unique()

# Build the category specific timeSeriesData
category = categories[0]
kts_category = build_kats_timeserie(dfp_data[dfp_data["category"] == category], "date", "num_sold")

# Train a model of forecast
model = build_model("theta", kts_category)
model.fit()

# Build predictions
dfp_forecast = model.predict(steps=365, freq="D")