The finest utility sector is agriculture, particularly in emerging nations like India. Utilizing historical data in agriculture may change the context of decision-making and increase farmer productivity. Approximately a part of India's population is employed in agriculture, however this sector contributes just 14% of India's GDP. This may be explained in part by farmers not making sufficient decisions on yield forecast. Increased agricultural yield is the outcome of accurate crop forecast. With this goal in mind, this work proposes the Improved Extreme Learning Machine (IELM) approach, which aims to forecast the best-yielding crop for a specific region by analyzing a variety of atmospheric factors, such as rainfall, temperature, humidity, etc., and land factors, such as soil pH and soil type, as well as historical data on crops grown. In this study, feature selection strategies are used to forecast crops using classification algorithms that recommend the best crop(s) for a given plot of land. After pre-processing the data to eliminate any undesirable information like NULL and other entries, this system is meant to forecast the best yield based on the dataset it has been given. Weak characteristics are eliminated using the Recursive Feature Elimination (RFE) feature selection approach until the necessary attributes are fulfilled. The IELM classifier beats the other learning strategies, according on the experimental findings.