Machine Learning for Integrated Farmer Advisory and Marketplace Systems: A Literature Survey

Farmers have to decide what to plant, what to buy, what to do when the crop goes wrong, and when to sell. Most of the software we found covers one of these decisions and leaves the rest alone. This paper looks at 27 studies, mostly from 2022 to 2026, on crop recommendation, fertilizer and input recommendation, advisory tools and disease detection, crop price prediction, online marketplaces, and a few joined-up platforms. For each one we note the method, the main finding, and what it leaves out if a farmer wants help from planting to sale. Tree-based ensembles turn up often in crop and input work, but price forecasting has no clear winner: ridge regression, LSTM and XGBoost each came out ahead in different studies. A handful of joined-up platforms exist, though they are mostly prototypes judged on benchmark data, and the e-NAM studies we read measure prices and arrivals but say nothing about advising the farmer. We also use our own prototype to show why accuracy figures need a simple baseline and real data before anyone leans on them. The paper ends with the gaps we found and a few things worth trying next.