Volume 2 - Issue 2, March - April 2026

This paper proposes a self-learning IoT-based system that leverages real-time soil and environmental sensor data to perform unsupervised analytics and adaptive crop recommendation. A distributed network of sensors (soil moisture, humidity, air-quality/gas, ammonia, and soil pH) collect high-frequency field data and transmit it through a low-power gateway to a cloud data platform. A data engineering pipeline handles ingestion, preprocessing, storage, and streaming analysis of this IoT data[1][2]. The core of the system applies unsupervised machine learning (clustering and anomaly detection) on the multivariate time-series data without relying on labeled datasets. Clustering algorithms (e.g. K-means) identify natural groupings of soil conditions, which are mapped to suitable crop types via agronomic heuristics[3][1]. In parallel, anomaly detectors flag outliers or sensor faults in real time, ensuring data quality and alerting to unusual field conditions[4][5]. We simulate sensor streams to demonstrate the pipeline: cluster scatterplots reveal distinct soil–environment regimes, and anomaly graphs show spikes being detected. Sample visualizations (e.g., sensor time series, cluster maps) illustrate how the system segments fields by moisture/pH profiles and provides crop recommendations per cluster. Experimental results (on synthetic data) confirm that the framework can autonomously learn from incoming data and suggest crops (such as rice, wheat, maize) suited to each soil cluster. The system thus provides a closed-loop decision support for precision farming without requiring historical labeled data.

Loganathan.S, Dr.Sreejith Vignesh B P,"A Self-Learning IOT-Based Soil and Environmental Intelligence System for Adaptive Crop Recommendation" International Journal of Advanced Multidisciplinary Research and Educational Development, V2(2): Page(465-470) Mar-Apr 2026. ISSN: 3107-6513. www.ijamred.com. Published by Scientific and Academic Research Publishing.