Enhancing Paddy Production Forecasts with Artificial Neural

Networks Models and Weather Data

M. Arjun, Eshwar Guntuka, M. Vara Prasad and Madhavi Devi B.

Department of Computer Science and Engineering, Institute of Aeronautical Engineering, Hyderabad, Telangana, India

Keywords: Rice Crop Yield Prediction, Artificial Neural Networks (ANN), Climate Data, Machine Learning, Multilayer

Perceptron, Agricultural Forecasting, Climatic Variability, Data Mining, Crop Modelling.

Abstract: Rice yield have a major role in ensuring no shortage of food, especially in countries like India, where a

significant portion of the population relies on rice as a staple food. However, rice yield is highly sensitive to

climatic variations, making accurate predictions essential to have effective- planning and making the

decisions. This paper presents an Artificial-Neural-Network (ANN) based model to predict rice crop yields,

using climatic data such as area, temperature (minimum, average, maximum), and Rainfall-of particular area.

The study focuses on 34 districts of Maharashtra, India, covering the Kharif season (between June and

November) for the years 1998 to 2019. The data was processed using the python libraries like numpy and

pandas’ tool, and a Multilayer Perceptron ANN model was developed with a feedforward training algorithm.

The ANN model demonstrated high prediction accuracy, achieving 97.54%, with sensitivity and specificity

of 96.33% and 98.12%, respectively. Additionally, performance metrics such as Mean-Absolute-Error (MAE)

(0.0526) and Root-Mean-Squared (RMS) Error (0.1527) indicate the robustness of the model. Compared to

traditional regression-based models, the ANN approach provided more accurate yield forecasts by effectively

capturing the nonlinear relationships between climatic variables and crop yield. This research contributes to

the ongoing efforts to enhance agricultural forecasting using AI techniques and offers a tool that can assist

farmers and policymakers in making informed decisions under variable climatic conditions. Future work aims

to expand the model by incorporating additional factors like soil quality and pest infestations, as well as real-

time data for dynamic predictions.

1 INTRODUCTION

Rice is an important and major food for a large part

of the global population and particularly in India,

where it is a major component of the diet. The

agricultural sector in India is highly dependent on rice

production, and the livelihoods of millions of farmers

are directly tied to the success of their crops.

However, due to the unpredictable and changing

nature of the climate, rice crop yields have shown

significant variability in recent years. This variability

poses a serious challenge to farmers, policymakers,

and the overall food supply chain.

Traditional crop yield prediction models, often

based on statistical methods like linear regression, are

limited in their ability to capture the intricate,

nonlinear relationship among climate variables (such

as rainfall, temperatures, and evapotranspiration) and

crop yields. These methods tend to oversimplify the

complex interactions that significantly impact crop

production as Ji et al., 2007 and Liu et al., 2001 have

demonstrated.

Artificial-Neural-Networks (ANNs), on a new

scenario, are well-suited to address this challenge.

With their ability to learn from data and capture

nonlinear relationships, ANNs offer a more accurate

and robust approach to predicting crop yields. This

paper introduces an ANN-based model developed for

predicting the rice crop yields in Maharashtra (state

of) India. By leveraging multiple climatic factors, the

model aims to provide more reliable predictions, will

be used assist farmers and stakeholders in making

informed decisions and mitigating the impacts of

climate change on agricultural productivity.

2 RELATED WORK

Various researches have explored the application of

machine learning methods to make crop yield

Arjun, M., Guntuka, E., Prasad, M. V. and B., M. D.

Enhancing Paddy Production Forecasts with Artiﬁcial Neural Networks Models and Weather Data.

DOI: 10.5220/0013889100004919

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies (ICRDICCT‘25 2025) - Volume 2, pages

731-734

ISBN: 978-989-758-777-1

731

prediction, having a particular focus on Artificial

Neural Networks (ANNs) as their ability includes to

model complex, nonlinear relationships. Often used

statistical techniques, such as linear regression, often

struggle to accurately predict crop yields, as they

oversimplify the interactions between climatic

variables like temperature, rainfall, and

evapotranspiration. These models are less effective in

handling the dynamic and nonlinear nature of

agricultural environments.

Recent advancements in ANN models have

demonstrated their superiority in predicting crop

yields over traditional regression methods. Ji et al.

(2007) highlighted the effectiveness of ANNs in

predicting rice yield, noting their ability to handle

complex environmental interactions, which

regression models often miss. Similarly, Smith et al.

(2009) applied ANN models to predict crop yields,

concluding that ANNs consistently outperformed

statistical models in diverse and fluctuating climatic

conditions. Their findings underscore the advantages

of neural networks in capturing the intricate

relationships between multiple climatic factors and

crop yields.

This research builds on the work of these studies

by applying ANNs to predict rice yields in

Maharashtra, India, using historical climatic data. By

focusing on variables such as temperature,

precipitation, and evapotranspiration, this study

focuses to further validate the accuracy of ANNs in

agricultural forecasting, particularly in regions with

variable climatic conditions. The proposed model is

designed to enhance prediction accuracy, supporting

better decision making for farmers and stakeholders.

3 PROPOSED METHODOLOGY

Using climatic and agronomic data, the work

estimates rice crop yield predicting with Levenberg-

Marquardt, Bayesian Regularization and Scaled

Conjugate Gradient algorithms. Features are of crop

year, area, rainfall, temperature, district and season

codes. The purposes of data normalization with

Standard-Scaler is to achieve mean and variance are

centered at zero, which improves the performance

and prediction results by the neural network.

3.1 Neural Network Architecture

This neural network used consists of an input layer

with six neurons (representing the six features),

single hidden layer with eight neurons, and a resultant

layer with a single neuron, which predicts the rice

production. The neurons in the network use the

Sigmoid-activation-function. The network is trained

using the LevenbergMarquardt (LM) algorithm,

which is ideal for nonlinear optimization problems

and converges faster than standard backpropagation

methods.

3.2 Flowchart for of ANN Algorithms

This following figure 1 flow-chart illustrates an

overall process for ANN algorithms for rice crop

yield prediction.

Figure 1: Flowchart of Overall for ANN Algorithm.

The training process for the network involves

minimizing the error among values of the predicted

and actual rice production using the ANN algorithms.

The LM algorithm is a combination of the Gauss-

Newton method and Gradient Descent, offering a

balance between speed and stability when converging

towards an optimal solution. The models were trained

with 500 epochs, ensuring that the network learns the

complex nonlinear relationships among the input

features, middle layers and rice production.

During testing phase, trained networks are used

for make predictions on the test data. Various metrics,

such as Mean-Squared-Error (MSE), Root-Mean-

Squared-Error (RMSE), Mean -Absolute-Error

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

732

(MAE), and Accuracy, are calculated for testing the

accuracy of the model. These metrics provide insights

into how well the model generalizes to new

information and its ability to predict rice yields

accurately.

This methodology provides a robust and efficient

way to predict rice crop yield, leveraging the LM

algorithm to capture the nonlinear relationships

among the input features middle layers and the target

variable. The neural networks architecture and

optimization ensure high accuracy and low error

rates, making this approach suitable for large-scale

crop yield prediction.

4 PRELIMINARY DATA

The dataset used to develop the Artificial- Neural-

Network (ANN) models consisted of climate and

agricultural data from Maharashtra, India, during the

Kharif season between 1998 and 2002. The features

included critical factors such as rainfall, minimum-

temperatures, average- temperatures, maximum-

temperatures, reference crop, area, and yield. The

data was pre-processed using the WEKA tool, where

missing values were handled, and features were

normalized to ensure a balanced input to the ANN

model. A multilayer perceptron architecture was

employed, with six input neurons corresponding to

the input variables, a hidden layer of three neurons,

and a single output neuron for predicting rice yield.

Dataset was divided for training and testing data, and

10-fold-cross validation was used for ensuring the

models generalizability and reliability among

different datasets of the data. Table 1 gives the

confusion matrix.

Table 1: A Confusion Matrix.

Predicted /

Observed

True False

True

True Positive

(

)

False Positive

(

)

False

False Negative

(FN)

True Negative

(TN)

5 EXPERIMENTAL RESULTS

The ANN model achieved remarkable prediction

performance with an overall accuracy of 97.54% This

research aligns with previous studies, such as those

conducted by Ye et al.2011, signifying its ability to

accurately forecast rice yields. The model's

sensitivity was 96.33%, meaning it correctly

predicted most positive cases, and its specificity was

98.12%, indicating that it effectively identified

negative cases. Performance metrics such as the

Mean -Absolute-Error (MAE), which stood at

0.0526, and the Root-Mean-Squared-Error (RMSE)

of 0.1527 further reinforced the model’s precision

and effectiveness. These findings show that the

ANN models highly reduced prediction errors when

comparing with traditional regression models,

making it a robust tool for rice yield forecasting in

Maharashtra. Figure 2 gives the neural network

diagram and Figure 3 gives the knowledge flow

layout.

Figure 2: Neural Network Diagram.

Figure 3: Knowledge Flow Layout.

6 DISCUSSION

Our model's high accuracy demonstrates the ANN's

ability to model nonlinear relationships between

climatic factors and rice yields. The results suggest

that climatic variables such as precipitation and

temperature are critical in determining crop yields,

and ANNs are well-suited to capture these

relationships. The model's success indicates the

potential for expanding its use to other regions and

crops. However, the model's dependence on historical

Enhancing Paddy Production Forecasts with Artiﬁcial Neural Networks Models and Weather Data

733

data highlights the need for updated, real-time

climatic data for more dynamic predictions.

7 CONCLUSIONS

This research helps finding the importance of ANN

models in predicting rice crop yields. By effectively

capturing the nonlinear interactions between climatic

variables, our model offers a robust alternative to

traditional prediction methods. The findings can

assist policymakers and farmers in making informed

decisions to mitigate the impacts of climatic

variability on rice production. Further work should

explore the integration of additional variables and

real-time data for more dynamic and precise

predictions.

REFERENCES

713–720, 1997.

Agricultural Science, 2016.

Anderson, R., Stenger, B., Cipolla, R., "Expressive Shape

Modeling Using Neural Networks," Artificial

Intelligence Review, vol. 43, pp. 155–177, 2012.

Batchelor, W.D., Yang, X.B., Tshanz, A.T., "Development

of a neural network for soybean rust epidemics,"

Transactions of the ASAE, vol. 40, pp. 247–252, 1997.

Dahikar, S., Rode, S., "Agricultural crop yield prediction

using artificial neural network approach," International

Journal of Innovative Research in Electrical,

Electronics, Instrumentation and Control Engineering,

vol. 2(1), pp. 683-686, 2014.

Elizondo, D.A., McClendon, R.W., Hoogenboom, G.,

"Neural network models for predicting flowering and

physiological maturity of soybean," Transactions of the

ASAE, vol. 37, pp. 981–988, 1994.

Gandhi, N., Armstrong, L., Petkar, O., "Predicting Rice

Crop Yield Using Bayesian Networks," Journal of

Gandhi, N., Petkar, O., Armstrong, L.J., "Rice Crop Yield

Prediction Using Artificial Neural Networks," IEEE

Govind, A., et al., "Machine learning-based rice yield

prediction model using remote sensing data," IEEE

Transactions on Geoscience and Remote Sensing,

2017.

International Conference on Technological Innovations in

ICT for Agriculture and Rural Development, 2016.

Jabjone, S., Wannasang, S., "Decision Support System

Using Artificial Neural Network to Predict Rice

Production in Phimai District, Thailand," International

Journal of Computer and Electrical Engineering, vol.

6(2), pp. 162-166, 2014.

Ji, B., et al., "Predicting rice yield based on environmental

factors using artificial neural networks," Journal of

Agricultural Science, vol. 145, pp. 249–261, 2007.

Ji, B., "Artificial neural networks for rice yield prediction

in mountainous regions," Journal of Agricultural

Science, vol. 145, pp. 249–261, 2007.

Ji, W., Cui, J., "Application of Geographical Information

System in Agricultural Land Classification and

Grading," Advances in Biomedical Engineering, vol.

3(5), pp. 201-205, 2011.

Kaul, M., Hill, R.L., Walthall, C., "Artificial neural

network for corn and soybean prediction," Agricultural

Systems, vol. 85(1), pp. 1-18, 2005.

Kaul, M., Hill, R.L., Walthall, C., "Artificial neural

network for corn and soybean prediction," Agricultural

Systems, vol. 85, pp. 1-18, 2005.

Liu, J., Goering, C., Tian, L., "A neural network for setting

target corn yields," Transactions of the ASAE, vol.

44(3), pp. 705-713, 2001.

Mohammadi, K., "Application of artificial neural networks

for predicting crop yield," Environmental Modeling &

Software, vol. 30, pp. 61-67, 2011.

O’Neal, M.R., Engel, B.A., Frankenberger, J., "Neural

network prediction of maize yield using alternative data

coding algorithms," Biosystems Engineering, vol.

83(1), pp. 31-45, 2002.

Pachepsky, Y.A., Timlin, D., Varallyay, G., "Artificial

neural networks to estimate soil water retention from

easily measurable data," Soil Science Society of

America Journal, vol. 60(3), pp. 727-733, 1996.

Puteh, S., et al., "Back propagation algorithm for rice yield

prediction," Ninth International Symposium on

Artificial Life and Robotics, Beppu, Japan, pp. 586-

589, 2004.

Ranjeet, T., Armstrong, L., "An architecture of a decision

support system for Western Australian Agriculture

Industry," 9th Conference of the Asian Federation for

Information Technology in Agriculture, Perth,

Australia, 2014.

Schaap, M.G., Bouten, W., "Modeling water retention

curves of sandy soils using neural networks," Water

Resources Research, vol. 32, pp. 3033–3040, 1996.

Smith, B.A., Hoogenboom, G., McClendon, R.W.,

"Artificial Neural Networks for Automated Year-

Round Temperature Prediction," Computers and

Electronics in Agriculture, vol. 68, pp. 52–6, 2009.

Starrett, S.K., Adams, G.L., "Using artificial neural

networks and regression to predict percentage of

applied nitrogen leached under turfgrass," Communica

tions in Soil Science and Plant Analysis, vol. 28, pp.

497–507, 1997.

Thomas, G.A., Taylor, G., Wood, J.C., "Mapping yield

potential with remote sensing," Precision Agriculture,

vol. 1, pp.

Yang, C.C., et al., "An artificial neural network model for

simulating pesticide concentrations in soil," Transactio

ns of the ASAE, vol. 40, pp. 1285–1294, 1997.

Ye, J., et al., "Neural network-based crop yield prediction

models for different agricultural systems," Computers

and Electronics in Agriculture, vol. 78, pp. 38– 50,

2011.

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

734