SynTSBench: A Synthetic Time Series Benchmark

SynTSBench is a comprehensive synthetic time series benchmark dataset designed for evaluating machine learning models on various time series tasks.

Dataset Structure

The dataset is organized into distinct configurations based on different time series characteristics and column structures:

• anomaly: Synthetic time series data from Dataset_generated_anomaly with columns: Feature1, date
• autocorrelation_long_dependency: Synthetic time series data from Dataset_generated_autocorrelation-long-dependency with columns: Feature1, date
• complex_multivariate_struct0: Synthetic time series data from Dataset_generated_complex_multivariate with columns: Feature1, Feature2, date
• complex_multivariate_struct1: Synthetic time series data from Dataset_generated_complex_multivariate with columns: consumer_spending, date, gdp_growth, inflation, interest_rate, unemployment_rate
• complex_multivariate_struct2: Synthetic time series data from Dataset_generated_complex_multivariate with columns: beverage_sales, date, ice_cream_sales, rainfall, temperature, umbrella_sales
• complex_multivariate_struct3: Synthetic time series data from Dataset_generated_complex_multivariate with columns: Feature1, Feature2, Feature3, date
• complex_multivariate_struct4: Synthetic time series data from Dataset_generated_complex_multivariate with columns: date, economic_growth, employment_rate, market_confidence, negative_indicator
• complex_multivariate_struct5: Synthetic time series data from Dataset_generated_complex_multivariate with columns: ad_spend, date, sales
• complex_multivariate_struct6: Synthetic time series data from Dataset_generated_complex_multivariate with columns: date, demand, price, supply
• complex_univariate: Synthetic time series data from Dataset_generated_complex_univariate with columns: Feature1, date
• cross_variable_learning_struct0: Synthetic time series data from Dataset_generated_cross-variable-learning with columns: FeatureA_WhiteNoise, FeatureB_Lag48, date
• cross_variable_learning_struct1: Synthetic time series data from Dataset_generated_cross-variable-learning with columns: FeatureA_WhiteNoise, FeatureB_Lag10, date
• cross_variable_learning_struct2: Synthetic time series data from Dataset_generated_cross-variable-learning with columns: FeatureA_Sin, FeatureB_Noise_0dB, FeatureC_Sum, date
• cross_variable_learning_struct3: Synthetic time series data from Dataset_generated_cross-variable-learning with columns: FeatureA_WhiteNoise, FeatureB_Lag5, date
• cross_variable_learning_struct4: Synthetic time series data from Dataset_generated_cross-variable-learning with columns: FeatureA_WhiteNoise, FeatureB_Lag24, date
• datasetlength: Synthetic time series data from Dataset_generated_datasetlength with columns: Feature1, date
• longdistance: Synthetic time series data from Dataset_generated_longdistance with columns: Feature1, date
• noise: Synthetic time series data from Dataset_generated_noise with columns: Feature1, date
• period: Synthetic time series data from Dataset_generated_period with columns: Feature1, date
• trend: Synthetic time series data from Dataset_generated_trend with columns: Feature1, date