haipipe-data-4-aidata

Preparing raw case data for machine learning model training involves transforming CaseSets into ML-ready AIDataSets by applying dataset splitting and input/output feature transformations. This process outputs HuggingFace DatasetDicts and vocabulary files required for training.

Splitting a patient dataset can be handled using strategies like RandomByPatient or time-based SplitByTimeBin. These methods divide raw case data into training, validation, and testing sets to ensure robust machine learning model evaluation.

Yes, you can transform time-series data into token sequences for feature engineering using the InputTEToken input transformation. This converts raw case features into token embeddings suitable for machine learning models.

Yes, the data pipeline explicitly supports generating HuggingFace DatasetDicts as an output. It transforms CaseSets into ML-ready AIDataSets and outputs these dictionaries alongside vocabulary files for model training.

Available data splitting strategies for machine learning datasets include time-based splitting via SplitByTimeBin, random splitting, and stratified splitting. These methods ensure proper partitioning into train, validation, and testing sets.

Converting raw case features into numerical sequences for ML models is achieved through input and output feature transformations. These transformations process CaseSets into ML-ready AIDataSets, generating formats like token embeddings for model training.