tsforecasting 1.2.55

TSForecasting is an automated time series forecasting framework

TSForecasting - Automated Time Series Forecasting Framework

Framework Contextualization

The TSForecasting project constitutes an complete and integrated pipeline to Automate Time Series Forecasting applications through the implementation of multivariate approaches integrating regression models referring to modules such as SKLearn, H2O.ai, XGBoost and also univariate approaches of more classics methods such as Prophet and AutoArima, this following an 'Expanding Window' performance evaluation.

The architecture design includes five main sections, these being: data preprocessing, feature engineering, hyperparameter optimization, forecast ensembling and forecasting method selection which are organized and customizable in a pipeline structure.

This project aims at providing the following application capabilities:

• General applicability on tabular datasets: The developed forecasting procedures are applicable on any data table associated with any Time Series Forecasting scopes, based on DateTime and Target columns to be predicted.

• Hyperparameter optimization and customization: It provides full configuration for each model hyperparameter through the customization of model_configs parameter variable values, allowing optimal performance to be obtained for each use case.

• Robustness and improvement of predictive results: The implementation of the TSForecasting pipeline aims to improve the predictive performance directly associated with the application of the best performing forecasting method.

Main Development Tools

Major frameworks used to built this project:

• Sklearn
• H2O.ai
• XGBoost
• AutoArima
• Prophet

Performance Evaluation Structure

The Expanding Window evaluation technique is a temporary approximation on the real value of the time series data. The first test segment is selected according to the train length and then it's forecasted in accordance with forecast size. The starting position of the subsequent segment is set in direct relation to the sliding window size, this meaning, if the window size is equal to the forecast size, each next segment starts at the end of the previous. This process is repeated until all time series data gets segmented and it uses all the iterations and observations to construct an aggregated and robust performance analysis to each predicted point.

Where to get it

The source code is currently hosted on GitHub at: https://github.com/TsLu1s/TSForecasting

Binary installer for the latest released version is available at the Python Package Index (PyPI).

Installation

To install this package from Pypi repository run the following command:

pip install tsforecasting

Usage Examples

1. TSForecasting - Automated Time Series Forecasting

The first needed step after importing the package is to load a dataset and define your DataTime and to be predicted Target column and rename them to 'Date' and 'y', respectively. The following step is to define your future running pipeline parameters variables, this being:

• train_size: Length of Train data in which will be applied the first Expanding Window iteration;
• forecast_size: Full length of test/future ahead predictions;
• window_size: Length of sliding window, Window_Size>=Forecast_Size is recommended;
• granularity: Valid interval of periods correlated to data -> 1m,30m,1h,1d,1wk,1mo (default='1d');
• eval_metric: Default predictive evaluation metric (eval_metric) is "MAE" (Mean Absolute Error), other options are "MAPE" (Mean Absolute Percentage Error) and "MSE" (Mean Squared Error);
• list_models: Select all the models intented do run in pred_performance function. To compare predictive performance of all available models set paramater list_models=['RandomForest','ExtraTrees','GBR', 'KNN','GeneralizedLR','XGBoost','H2O_AutoML','AutoArima','Prophet'];
• model_configs: Nested dictionary in which are contained all models and specific hyperparameters configurations. Feel free to customize each model as you see fit (customization example shown bellow);

The pred_performance function compares all segmented windows values (predicted and real) for each selected and configurated model then calculates it's predicted performance error metrics, returning the variable best_model(most effective model), perf_results[DataFrame] containing every detailed measure of each Test predicted value and at last the variable predictions[DataFrame] containing every segmented window iteration performed which can be use for analysis and objective models comparison.

The pred_results function forecasts the future values based on the previously predefined parameters and the selected model which specifies the choosen model used to obtain future predictions.

Importante Note:

• Although not advisable to forecast without evaluating predictive performance first, forecast can be done without using the pred_performance evaluation function, by replacing the selected_model parameter (default='RandomForest') in the pred_results function with any choosen model.

import tsforecasting as tsf
import pandas as pd
import warnings
warnings.filterwarnings("ignore", category=Warning) #-> For a clean console
import h2o

h2o.init() # -> Run only if using H2O_AutoML models   

data = pd.read_csv('csv_directory_path') # Dataframe Loading Example
    
data = data.rename(columns={'DateTime_Column': 'Date','Target_Name_Column':'y'})
data=data[['Date',"y"]]
    
train_size_=0.95
forecast_size_=15
window_size_=Forecast_Size # Recommended
granularity_='1d' # 1m,30m,1h,1d,1wk,1mo
eval_metric_="MAE" # MAPE, MSE
list_models_=['RandomForest','ExtraTrees','KNN','XGBoost','AutoArima'] # Ensemble Example
    
## Get models hyperparameters configurations
    
models_hparameters=tsf.model_configurations()
print(models_hparameters)

# Customization Example
models_hparameters["RandomForest"]["n_estimators"]=100
models_hparameters["ExtraTrees"]["n_estimators"]=100

## Performance Evaluation
best_model,perf_results,predictions=tsf.pred_performance(Dataset=data,
                                                         train_size=train_size_,
                                                         forecast_size=forecast_size_,
                                                         window_size=window_size_,
                                                         list_models=list_models_,
                                                         model_configs=models_hparameters,
                                                         granularity=granularity_,
                                                         eval_metric=eval_metric_)
## Forecast
dataset_pred=tsf.pred_results(Dataset=data,
                              forecast_size=forecast_size_,
                              model_configs=models_hparameters,
                              granularity=granularity_,
                              selected_model=best_model)

2. TSForecasting - Extra Auxiliar Functions

The model_prediction function predicts your Test target column based on the input DataFrames, Train and Test, model configuration set by the parameter model_configs and the selected running algorithm in the parameter algo (default='RandomForest'). Note, you can select and customize any of the 11 models available in Model_Configs dictionary.

 
# Automated Model Predictions
 
y_predict = tsf.model_prediction(Train:pd.DataFrame,
                                 Test:pd.DataFrame,
                                 target:str="y",
                                 model_configs:dict=models_hparameters,
                                 algo:str='RandomForest')

The engin_date function converts and transforms columns of Datetime type into additional columns (Year, Day of the Year, Season, Month, Day of the month, Day of the week, Weekend, Hour, Minute) which will be added by association to the input dataset and subsequently deletes the original column if variable Drop=True.

The multivariable_lag function creats all the past lags automatically (in accordance to range_lags parameter) and adds each column into the input DataFrame.

# Feature Engineering 
    
dataset = tsf.engin_date(Dataset:pd.DataFrame,
                         Drop:bool=False) 

dataset = tsf.multivariable_lag(Dataset:pd.DataFrame,
                                target:str="y",
                                range_lags:list=[1,10],
                                drop_na:bool=True)
    

This feature_selection_tb function filters the most valuable features from the dataset. It's based on calculated variable importance in tree based regression models from Scikit-Learn and it can be customized by use of the parameter total_vi (total sum of relative variable\feature importance percentage selected) and algo selecting the model for evaluation ('ExtraTrees','RandomForest' and 'GBR').

# Feature Selection 

selected_columns, selected_importance_df=tsf.feature_selection_tb(Dataset:pd.DataFrame,
                                                                  target:str="y",
                                                                  total_vi:float=0.99,
                                                                  algo:str="ExtraTrees",
                                                                  estimators:int=250)

You can analyse the obtained performance results by using the metrics_regression function wich contains the most used metrics for regression predictive contexts.

 
# Regression Performance Metrics

reg_performance = pd.DataFrame(tsf.metrics_regression(y_true,y_pred),index=[0])    # y_true:list, y_pred:list
        

License

Distributed under the MIT License. See LICENSE for more information.

Contact

Luis Santos - LinkedIn

Feel free to contact me and share your feedback.