Computer-implemented method for predicting future data values and a predictive system

PL452032A1Pending Publication Date: 2026-11-16BIZ ON SPÓŁKA Z OGRANICZONĄ ODPOWIEDZIALNOŚCIĄ
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
PL2025452032
Authority / Receiving Office
PL · PL
Patent Type
Applications
Current Assignee / Owner
Filing Date
2025-05-13
Publication Date
2026-11-16

Smart Images

  • Figure P_ABST
    Figure P_ABST
Patent Text Reader

Abstract

A computer-implemented method for predicting future values ​​of time-varying data, in which an output prediction (OUT) is generated based on input data (IN) using a machine learning module (ML), wherein the input data (IN) and the prediction (OUT) are stored in a database (BD), is characterized in that it uses data with at least two frequencies, which are stored in a cache (CACHE), and the data with a lower frequency are used to determine the prediction, while the data with a higher frequency are used to validate the prediction using a deep learning module (DL). The output prediction is issued provided that the prediction obtained using the module (ML) is positively consistent with the validation obtained using the module (DL). A record containing the input data, the prediction result, and the validation result is stored in a database (BD). The database (BD) uses a storage based on NVMe semiconductor non-volatile memory.Before validation using the deep learning (DL) module, it is checked whether a record containing such input data has been stored in the database (DB). Validation using the deep learning (DL) module is performed only if such a record does not exist, otherwise returning a result consistent with the stored record. The subject of the application also includes a predictive system for implementing the method.
Need to check novelty before this filing date? Find Prior Art