Abstract
Measuring soil moisture has a big impact on resource efficiency and decision-making in geotechnical engineering, agriculture, and environmental sustainability. This study presents a revolutionary Internet of Things-based method for predicting soil type and moisture in real time. Water content was verified using the traditional oven-dry method for accuracy evaluation, and capacitance measurements were gathered using a sensor to generate a custom dataset in the lab. The method outperformed linear regression in water content prediction, achieving 96.49% accuracy using an Excel logarithmic regression equation. Furthermore, a machine learning model that employed polynomial regression was able to measure the water content of the soil and predict values for fine, medium-coarse, and coarse sand types. With an R2 score of 0.79, the model can account for almost 79% of the variation in water content and produce 1.71% of MAE(Mean Absolute of Error), indicating a strong relationship between capacitance and water content. On the expanded dataset, Random Forest classifier was chosen for classification, which correctly identified the intended soil types with an accuracy of roughly 97.77%. By combining sensor data with sophisticated algorithms, the suggested methodology makes it possible to analyze soil qualities effectively and non-destructively. This scalable method offers substantial potential for environmental management, soil monitoring, and precision agriculture and is flexible enough for the creation of mobile applications. Predictive modeling and real-time data processing combined improve resource management effectiveness while lowering the need for human intervention. In order to improve forecast accuracy and application and support more environmentally friendly farming methods and environmental monitoring, further research will surely give efforts to enrich the dataset, incorporate a variety of soil types, and take environmental elements like temperature and salinity into consideration.