Arquitectura de oráculo blockchain con validación de datos basada en procesos de aprendizaje de maquina

The growing reliance on smart contracts in blockchain ecosystems has highlighted the need for reliable mechanisms to validate the data these contracts consume from external sources, commonly managed by oracles. This article presents a blockchain oracle architecture that incorporates a data validation system based on machine learning techniques, with the goal of enhancing data quality before it is incorporated into the blockchain. The solution integrates supervised and unsupervised models for anomaly detection, multi-label classification, time series analysis, and outlier detection. The validation process follows the CRISP-DM methodology and is complemented by a data integrity indicator based on descriptive statistics and majority voting mechanisms (hard voting), which allows for the automatic estimation of data acceptability. Experimental results, obtained in a testing environment using functional smart contracts, demonstrate improvements in the detection of inconsistencies and data manipulation, as well as in the reliability of automated decisions. This proposal provides a systematic and replicable strategy to mitigate risks associated with data consumption in blockchain applications.