Prediction model to select best fertilizer for the harvested crop
India has a large agricultural sector supporting majority of the population for their livelihood. Every year huge resources in terms of land and fertilizer are used for production of food. Most of the fertilizers being used in farms today are being wasted due to the incorrect use of type of fertilizer. Farmers do not have enough of knowledge about soil nutrients contents such as nitrogen, phosphorous and potassium in the land. However, the balanced dose of fertilizer used for different crop is very important for obtaining desirable yield which depends on several edaphic and environmental factors. In addition, the elemental composition especially nitrogen, phosphorus and potassium content of a fertilizer needs to be considered in selecting a source fertilizer as these elements play a vital role in plant growth and development. In this regard, a fertilizer requirement prediction model would be very useful for obtaining higher yield of different crops in India.
In this present project, we will use the standard machine learning models that will analyze various soil features like N, P, K, soil pH, and environmental parameters such as temperature, relative humidity, and annual rainfall of particular land area to recommend the type of fertilizer to be used for the selected crops. Therefore, the present project has been proposed to analyze available datasets and to predict fertilizer requirements for two different crops based on soil features and environmental parameters using appropriate machine learning models.
In this dataset, target feature is labeled and categorical, so we decide to use supervised classification algorithms like K-Nearest Neighbors, Decision Tree, Random Forest, and Gradient Boosting. For classification and machine learning we will use Pandas, NumPy, Scikit-learn, Matplotlib and Seaborn libraries. The following steps will be followed to achieve the goal of this project namely- i) Exploratory Data Analysis for visualizing data, ii) Data preprocessing, iii) Divide the dataset into train and test set to fit into classification models. iv) Performance evaluation of models with accuracy score, Recall, Precision and F1-score and v) Prediction analysis.
From this project we will be able to find a suitable machine learning model to predict fertilizer that best suits a particular type of soil and crop based on other environmental consequences. Our proposed machine learning model will help growers to decide appropriate fertilizer rapidly and economically to reduce fertilizer losses.