truesight 0.0.2a2

Truesight is a python package for time series prediction using deep learning and statistical models.

TrueSight ✨

The TrueSight model is a hybrid forecasting tool that uses statistical forecasting models together with a Deep Neural Network (DNN) to make predictions. The TrueSight Preprocessor class is responsible for getting all the statistical forecasters in one place. It can handle forecasters from packages like statsforecast, scikit-learn, pmdarima, and others. You just need a class that accepts the seasonal_length parameter for the constructor and has a .fit(x, y) method

All you need to do before using this package, is create a pandas dataframe with the following structure:

• unique_id: A string that uniquely identifies each time series in the dataframe
• ds: A datetime column with the date of each time step. The dates must be in the correct frequency for the date_freq parameter
• y: The values of the time series

and run the steps in the Usage section. Easy peasy! 😎

Instalation 💻

To install the TrueSight package, just run:

pip install truesight

We also recommend installing the statsforecast package for the statistical forecasters:

pip install statsforecast

Usage 🚀

Import the necessary modules

import tensorflow as tf
from truesight.preprocessing import Preprocessor
from truesight.core import TrueSight
from truesight.metrics import Evaluator, smape, mape, mse, rmse, mae
from truesight.utils import get_input_shapes, generate_syntetic_data

Load the data

num_time_steps = 60
seasonal_lenght = 12
forecast_horizon = 12
df = generate_syntetic_data(num_time_steps, seasonal_lenght, 100)

Create and run the preprocessor class. You can include as many statistical models as you need in the models parameter. Just make sure they follow the syntax similar to statsforecast. However, keep in mind that more models mean longer processing time. It's important to set a fallback model in case any of the informed models fail to fit.

from statsforecast.models import SeasonalNaive, AutoETS
from truesight.models import AdditiveDecomposition

preprocessor = Preprocessor(df)
X_train, Y_train, ids_train, X_val, Y_val, ids_val, models = preprocessor.make_dataset(
    forecast_horizon = 12, 
    season_length = 12,
    date_freq = "MS", 
    models = [AdditiveDecomposition, AutoETS, SeasonalNaive], 
    fallback_model = SeasonalNaive,
    verbose = True
    )

Create the model

input_shapes = get_input_shapes(X_train)
truesight = TrueSight(models, input_shapes, forecast_horizon = forecast_horizon)
truesight.auto_tune(X_train, Y_train, X_val, Y_val, n_trials = 50, batch_size = 512, epochs = 5)

Use the auto_tune to automatically define the hyperparameters

truesight.auto_tune(X_train, Y_train, X_val, Y_val, n_trials = 10, batch_size = 512, epochs = 5)

Or set then manually

truesight.set_hparams(lstm_units=256, hidden_size=1024, num_heads=8, dropout_rate=0.1)

Train the model, as the model is built on the tensorflow framework, any tensorflow callback can be used

callbacks = [
    tf.keras.callbacks.EarlyStopping(patience = 100, restore_best_weights = True, monitor = "val_loss"),
    tf.keras.callbacks.ReduceLROnPlateau(monitor = "val_loss", factor = 0.5, patience = 25, verbose = 1),
]
truesight.fit(
    X_train, Y_train, 
    X_val, Y_val, 
    batch_size = 128, 
    epochs = 1000, 
    verbose = False, 
    callbacks = callbacks,
 )
truesight.plot_history()

Evaluate the results

Y_hat = truesight.predict(
    X_val, 
    batch_size = 500, 
    n_repeats = 100, 
    n_quantiles = 15, 
    return_quantiles = True, 
    verbose = False,
 )
evaluator = Evaluator(X_val, Y_val, Y_hat, ids_val)
evaluator.evaluate_prediction([smape, mape, mse, rmse, mae], return_mean=False)

id	mse	rmse	mae
10	1164.51	34.1249	25.9397
13	2.91094	1.70615	1.69313
18	2.62309	1.6196	1.5549
38	2.77819	1.66679	1.62896
45	892.701	29.8781	10.0897
46	2.96284	1.72129	1.71549
50	3585.54	59.8794	49.0362
51	12199.1	110.449	93.9491
53	2.6345	1.62311	1.56625
54	2.77184	1.66488	1.63493
61	303.464	17.4202	14.6694
67	2.70393	1.64436	1.60988
69	2578.09	50.7749	38.986
70	3.09074	1.75805	1.75805
73	2.80945	1.67614	1.65109
75	2.88333	1.69804	1.68085
78	3.09091	1.7581	1.7581
81	2.64539	1.62647	1.56712
83	491.263	22.1644	18.4432
86	2.84002	1.68523	1.64177

evaluator.plot_exemple()