Long-term time series forecasting is important in energy scheduling, traffic management, meteorological monitoring, and industrial operations. However, conventional statistical models and recurrent neural networks have limitations in modeling long-range dependencies, complex periodic patterns, and multivariate relationships. This review compares three representative Transformer-based models—Informer, Autoformer, and PatchTST—through literature synthesis and comparative analysis. Informer reduces long-sequence computation through sparse attention, Autoformer strengthens periodic modeling through series decomposition and Auto-Correlation, and PatchTST improves input representation through patch-based tokenization. Public results on the Electricity dataset show that PatchTST outperforms earlier Transformer-based models, while DLinear achieves comparable errors. These findings indicate that model complexity is not the sole determinant of forecasting performance and that look-back windows and experimental protocols also affect model comparisons. The review further discusses computational cost, non-stationarity, and cross-variable modeling.

Show more

With the increasing complexity of urban traffic systems, reliable forecasting of road conditions has become an important requirement for congestion management and intelligent transportation applications. This study develops a multivariate and multi-horizon traffic speed forecasting framework using real-world traffic monitoring data collected from 75 street segments in Shenzhen, China. LSTM and PatchTST are selected as representative recurrent and Transformer-based forecasting models for comparative evaluation. Average travel speed is considered the prediction target, while traffic index, total sample travel length, and total sample travel time are incorporated as additional input variables. Historical observations from the previous 48 hours are used to forecast traffic speeds over three future horizons, including 30, 60, and 120 minutes. To increase the dependability of experimental outcomes, the dataset is split chronologically, and several random seeds are used. MAE, RMSE, MAPE, and R2 are used to assess the model's performance. According to the experimental findings, PatchTST consistently outperforms LSTM in predicting across all prediction horizons. The MAE values of PatchTST are 2.0710, 2.1960, and 2.3552 for 30-, 60-, and 120-minute forecasting tasks, corresponding to reductions of 6.81%, 7.86%, and 5.89% compared with LSTM. Meanwhile, PatchTST obtains R² values of 0.8886, 0.8795, and 0.8670, respectively. Although the forecasting accuracy of both models decreases with longer prediction horizons, PatchTST maintains relatively lower errors and better stability. The results indicate that the patch-based representation and attention mechanism provide advantages in capturing both local variations and long-term temporal dependencies in urban traffic sequences.

Show more

This study examines the temporal and spatial patterns of Toronto Transit Commission (TTC) subway delays in Toronto using the TTC Subway Delay Data. The dataset contains 28,191 delay records from January 1, 2025, to January 31, 2026, and the 2025 records are selected for analysis. Descriptive statistics and data visualization are used to examine hourly and station-level delay patterns. A random forest classification model is developed to predict whether at least one delay incident will be recorded during a specific date-hour period at Union Station. The results show that TTC subway delays are unevenly distributed across time and stations. Delay incidents are more frequent around 22:00, while several major stations, including the Kennedy Bloor-Danforth Line (BD), Bloor, Finch, Kipling, and Union Station, record relatively high numbers of delay incidents. The classification model achieves relatively high recall but low precision, indicating that it identifies many date-hour periods containing recorded delay incidents.

Show more

Traditional wine quality assessment based on tasters is costly and subjectively inconsistent, making it difficult to meet the needs of automated quality control for large volumes of data. This study imports the 1,143-record WineQT dataset into MySQL, checks data quality, and constructs two SQL-derived features before six-class modeling. The initial row-wise split gave the full-feature random forest an accuracy of 0.7118 and a macro-F1 of 0.3415. A later audit, however, identified 125 repeated rows. Among these, 43 test records had identical counterparts in the training set. Keeping identical observations in the same group reduced accuracy to 0.5721 and macro-F1 to 0.2713, providing a more conservative estimate. In repeated stratified group five-fold cross-validation, the original features achieved 0.5984±0.0300 accuracy and 0.2793±0.0218 macro-F1; adding the SQL-derived features produced no improvement. Class weighting raised macro-F1 but lowered accuracy and weighted-F1. SQL therefore contributes mainly through consistent data checks, feature definitions, and a more traceable evaluation process.

Show more