climate, and market demands, as well as a lack of proper guidance. A proposed AI-based crop

recommendation system that helps the farmers take proper decisions in the field of agriculture. The

proposed system uses machine learning algorithms to consider important factors like soil nutrients,

temperature, humidity, rainfall, pH, and current market demands to recommend the best crop for farming.

A multilingual web-based system helps the farmers of different linguistic backgrounds to input the

environmental values and receive the recommendations in the desired language. The proposed system helps

to increase the yield, optimize resources, and maximize the returns while reducing the risk of crop failure.

The results of the experiments show the accuracy and efficiency of the proposed model.

Index Terms: Crop Recommendation System, Precision Agriculture, Soil Nutrients Analysis, Agriculture

analytics

become less reliable. Therefore, farmers often experience difficulties in crop production, economic losses,

and inefficient utilization of resources. Today, with significant improvements in Artificial Intelligence (AI)

and Machine Learning (ML), there is an opportunity to transform traditional agriculture into more

intelligent agriculture. AI-based solutions have been capable of handling huge amounts of agricultural data

and identifying patterns that were not easily recognizable using traditional techniques.

The proposed AI-based crop recommendation system helps overcome these challenges by incorporating

machine learning and agricultural domain knowledge. The system can collect input parameters like soil

nutrients, climatic conditions, and market trends, and then use machine learning algorithms to predict the

most suitable and profitable crops for farming.

Machine learning algorithms, such as supervised learning algorithms help identify complex patterns and

the farmers and the system, making it easier for them to use the system. The proposed system helps

overcome the limitations of traditional guesswork and promotes informed decision-making. It

optimizes resource utilization, minimizes the risk of crop failures, and maximizes agricultural productivity.

It also helps implement sustainable farming by promoting the growth of crops that are most suitable for

the environment. The adoption of such AI-based systems also provides certain benefits. It reduces

dependency on guesswork and traditional practices. It also reduces the risk of crop failure. Moreover, it

encourages the efficient utilization of resources. The system provides recommendations on which crop is

most suitable for cultivation in the given conditions. Such recommendations also increase crop yields.

Furthermore, it encourages sustainable agricultural practices. It encourages the optimal utilization of

natural resources. It also reduces environmental impacts. However, there are certain challenges associated

The integration of Artificial Intelligence (AI), Machine Learning (ML), and Internet of Things (IoT) has

greatly impacted the modernization of conventional farming practices. Over the last decade, many research

papers have been published to investigate different computing techniques for crop recommendation, yield

prediction, irrigation systems, and smart farming. This has been motivated by the need to address some of

the most critical problems affecting modern agriculture, including climate change, soil erosion, and

inefficient use of resources.

Previous research has been focused on developing crop recommendation systems using machine learning

techniques. Machine learning algorithms were primarily considered for developing crop recommendation

systems using soil and environmental parameters. Apat et al. proposed an artificial intelligence-based crop

machine learning algorithms in processing environmental and soil-related parameters. This system is

effective in helping the farmer make the best choice of crop.

In reference [3], Bali and Singla performed an in-depth survey of machine learning techniques in crop yield

prediction. The survey highlighted different algorithms, such as decision trees, support vector machines,

and ensemble methods, and their performance in agricultural applications. The survey also highlighted the

significance of feature selection and data preprocessing in improving the accuracy of the models. In reference

[4], Sinha and Dhanalakshmi performed an in-depth survey of the latest developments in IoT-based smart

agricultural systems. The survey highlighted the importance of IoT devices in monitoring environmental

parameters such as soil moisture, temperature, and humidity. The survey also highlighted the significance of

real-time data collection in improving the accuracy of the models. In reference [5], Qazi et al. performed an

proposed a big data-driven system in cultivation planning. The system assisted farmers during the

preliminary stages of production. The survey highlighted the significance of data-driven decision-making in

improving agricultural outcomes. Nandhini et al. in [8] suggested the idea of improving crop health by

making use of advanced deep learning techniques with the help of a neural network. The research makes use

of skewness fully connected neural networks for the purpose of data analysis and improving crop cultivation.

The paper deals with different tools and techniques related to deep learning and their opportunities in smart

farming. It also explains the capabilities of neural networks in dealing with large amounts of data in the field

of agriculture and making accurate predictions. In [10], Nevavuori et al. have used deep learning based

convolutional neural networks in crop yield prediction. The paper proves the capabilities of CNN in dealing

with large amounts of data in the field of agriculture and making accurate predictions when compared to

deep learning in the field of smart farming and its capabilities in enhancing productivity and efficiency in

the field of agriculture. In [12], Farooq et al. have given an exhaustive account of the role of IoT in the field

of agriculture and its application in implementing smart farming. The paper deals with different IoT

architectures and their capabilities in enhancing the productivity of the field of agriculture. In [13], Gayatri

et al. suggested an IoT-based smart agriculture system that offers real-time solutions for farmers. The system

offers the integration of IoT and communication technologies to monitor crop conditions and optimize crop

prediction in reducing uncertainties and improving crop yield. The study also emphasizes the significance

of AI in improving the accuracy of environmental predictions. Despite all these developments, there are

some challenges associated with the existing systems. Some of the challenges include the fact that most of

the models used for crop recommendation are static, meaning they do not adapt to changes in real-time.

Moreover, some of the models are complex, making them difficult for farmers with little knowledge of how

to use them. Finally, most of the models focus only on environmental factors, with no attention to economic

factors such as market trends and profitability of crops. Moreover, there are challenges associated with the

availability of data, which affect the performance of machine learning models in agricultural systems. The

proposed system overcomes the above-mentioned

integrating

machine learning with real-

The proposed AI-based crop recommendation system is expected to assist farmers in identifying the best

and most profitable crops to cultivate based on certain environmental, soil, and market conditions. The

proposed system is based on a modular framework comprising multiple components that interact with each

other to provide farmers with precise and real-time crop recommendations.

The user interface is an interactive platform through which farmers can interact with the proposed AI-based

crop recommendation system. In this context, the user interface is developed using React, which is used to

build web applications. In other words, it is a web-based application that can be used by farmers to input

various agricultural parameters. In this context, farmers can input data such as nutrients in the soil (Nitrogen,

Phosphorus, Potassium), temperature, humidity, rainfall, pH, and market conditions. In addition, the

words, it is possible to collect data from various sources and analyze it in real-time. In addition, data

preprocessing techniques are applied to input data. In addition, feature selection is also performed to identify

the most significant parameters affecting crop growth.

The core of the system is the machine learning module. This module is responsible for making crop

predictions. The system uses supervised learning techniques, specifically MLP Classifier, TabTransformer,

and TabNet. These techniques have been proven effective in making accurate predictions on complex data.

The system is trained on agricultural data that contains information on soil conditions, climatic conditions,

and crop types. Once trained, the system analyzes the input data and makes accurate predictions.

Performance and Evaluation

The performance of the proposed AI-based crop recommendation system is evaluated using different

metrics and visualization techniques. The evaluation of the proposed system is based on the accuracy,

learning characteristics, and prediction ability of the implemented machine learning models, namely MLP

Classifier and TabTransformer.

Model Training Performance

The accuracy and validation accuracy of the models are examined to assess the learning behavior of the

models. The accuracy of the implemented models is found to be high, as depicted in Figure 4. The

accuracy of the implemented models is found to be 98% during training, approaching almost optimal

accuracy. The accuracy of the implemented MLP model is found to increase in every epoch,

implemented models are effective in generating accurate predictions through effective feature selection

and model selection.

Feature Importance Analysis

Feature importance is analyzed for the implemented models. The feature importance is analyzed for the

implemented MLP Classifier model, as depicted in Figure 5. The feature importance is found to be higher

for environmental factors like humidity and rainfall. The feature importance is also found to be higher for

soil nutrients like Nitrogen (N), Phosphorus (P), and Potassium (K).

Confusion Matrix Evaluation

The confusion matrix of the TabTransformer model, which evaluates the classification performance of the

proposed system for different types of crops. The diagonal values of the matrix denote the accurate

classification of crops, whereas the non-diagonal values denote the classification errors. From the matrix, it

can be observed that there are a significant number of diagonal values, proving that the model accurately

classifies most of the crops. The presence of a small number of non-diagonal values in the matrix proves

that there are minimal classification errors, ensuring precision in distinguishing different types of crops.

The results of this analysis prove the effectiveness and reliability of the proposed model, implying that the

accuracy of the model for most of the classes is high. This shows that the model has high precision and

consistency in its predictions regardless of the conditions. The matrix, therefore, implies that the model can

quality of predictions. The proposed system allows for real-time predictions with minimal time lag.

From the evaluation results, it can be concluded that the proposed system helps the farmers make informed

decisions, resulting in the growth of crops..