Forecasting hourly with Prophet

Question

I have hourly humidity data of a place from 2017 - 2022. So, I use 2017 - 2021 data as a training data to forecast the hourly humidity for 2022, and compare the result to the true humidity from the data that I have.

I use Prophet with no arguments:

model = new Prophet()

I got MAE of 8.57 and here I plot the daily (resampled) forecast of y, yhat, upper and lower yhat.

I'm new to this time series forecasting and I'd like some advice about my case. My question is, what can I do to improve this forecast?

Thanks!

Answer 1

For improving models performance you need to do hyperparameter tuning. These are the parameters that significantly impact the output of you model. For tuning your hyperparameters, you need to do cross validation followed by best parameter selection. Docs

import itertools
import numpy as np
import pandas as pd
from prophet.diagnostics import performance_metrics, cross_validation 

from prophet import Prophet


param_grid = {  
    'changepoint_prior_scale': [0.001, 0.01, 0.1, 0.5],
    'seasonality_prior_scale': [0.01, 0.1, 1.0, 10.0],
}

# Generate all combinations of parameters
all_params = [dict(zip(param_grid.keys(), v)) for v in itertools.product(*param_grid.values())]
rmses = []  # Store the RMSEs for each params here

# Use cross validation to evaluate all parameters
for params in all_params:
    m = Prophet(**params).fit(df)  # Fit model with given params
    df_cv = cross_validation(m, cutoffs=cutoffs, horizon='30 days', parallel="processes")
    df_p = performance_metrics(df_cv, rolling_window=1)
    maes.append(df_p['mae'].values[0])

# Find the best parameters
tuning_results = pd.DataFrame(all_params)
tuning_results['mae'] = maes
best_params = all_params[np.argmin(maes)]
print(best_params)
better_model = Prophet(**best_params)