A Resource-Constrained AI Factory Framework with Data Schema and Validation for Satellite-Based Agricultural Monitoring in Pakistan

Introduction/Importance of Study:Pakistan’s agriculture is highly dependent on crop monitoring;existing solutions are limited due to fragmented data, infrastructural issues, and the lack of analytical solutions, affecting the use of satellite-based data for efficient agricultural decision-making.Novelty Statement:This research proposes a novel solution for an AI factory framework designed for resource-constrained environments. It incorporates a structured data pipeline, Quality Data Schema (QDS), and Quality Validation System (QVS) for efficient and reliable agricultural monitoring applications.Material and Method:A synthetic dataset was developed to simulate vegetation indicesand environmental factors in a noisy and uncertain environment. This research proposes a structured data pipeline solution involvingdata preprocessing, structured data handling, and multi-level validation, along with a random forest model for efficient results.Result and Discussion:The experimental results prove an improvement in accuracy from 82.67% to 83.33% and F1-score from 0.82 to 0.83. The statistical results over 20 runs show a mean accuracy of 0.8242 ± 0.0136, and the results for accuracy have a confidence interval of ±0.0059. The paired t-test results indicate that these improvements are not statistically significant at a 95% confidence level. The key contribution is in improving system reliability.Concluding Remarks:The experimental results showed that the structured data handling and validation within an AI factory framework improve data reliability, which is useful for deploying AI-based agricultural monitoring systems in Pakistan.