Anonymized Multimodal Dataset for Urban Traffic Prediction: CDMX

This dataset provides 300,000 anonymized and preprocessed records for urban traffic forecasting using deep learning. Derived from TomTom NV raw data, the features (X1 ... X20) cover historical speeds, road geometry, weather conditions, and cyclical temporal components. All variables are statistically standardized (Z-Score and Min-Max) to ensure model stability. The target variable (y) is a variational stabilization of the Relative Velocity Index (RVI) through a logarithmic transformation ln(RVI+1). This resource is intended to ensure the reproducibility of the doctoral research conducted at CENIDET regarding kinematic normalization in urban corridors.