Quantile correlative deep feedforward multilayer perceptron for crop yield prediction☆
Introduction
Deep learning is employed in the development of decision support systems in different domains. Deep Learning is used to apply in other important domains such as agriculture. Agriculture plays a significant task in the Indian economy since it provides large employment opportunities. Agricultural yield is mainly used in India based on weather conditions. The country's economy mainly depends on agriculture for rural sustainability. Crop yield prediction is a significant agricultural issue. Crop yield is termed a key indicator of sustainable growth in agriculture. Early and accurate information on crop growth conditions is used for estimating crop yields. Crop yield prediction is an essential one in agriculture to recognize the income of farmers. Agricultural yield depends on the weather conditions, pests, and planning of harvest operation. The crop yield prediction is one of the most attractive challenging tasks in the agriculture sector. Recently, several machine learning and deep learning methods have been developed for crop yield prediction. A novel end-to-end prediction model that combines two back-propagation neural networks (BPNNs) with an independently recurrent neural network (IndRNN), called BBI-model was developed [1] for rice yield prediction. The designed BBI-model achieved the best prediction performance with minimum mean absolute error. The data preprocessing stage was used to the original data with aid of missing-value interpolation and data normalization technologies. The original area data and meteorological data into numerical data were converted into time-series data with help of a one-hot code. One BPNN and the IndRNN were used to create the deep spatial and temporal features for learning the interaction among them and the summer and winter rice yields. But the model failed to focus on the important influence of other features to improve prediction performance.
A Modular Artificial Neural Networks and Support vector regression (MANNs-SVR) was developed [2] Crop yield prediction based on aggregated rainfall data. The designed model accurately predicts the Kharif crop production. Modular artificial neural networks (MNNs) were applied for predicting the amount of rainfall that occurs in the monsoon season in Visakhapatnam. After predicting rainfall, feature selection was used for choosing only significant features in predicting different Kharif crops in Visakhapatnam. Lastly, the Kharif crops were correctly predicted. The predictions only considered the area and rainfall but they failed to consider the other climatic parameters such as temperature to improve the performance of Kharif crop yield prediction. A modular and reusable machine learning workflow was developed [3] for crop yield prediction. The designed model minimizes the error rate of crop yield prediction. But the deep learning technique failed to apply for further improving the accuracy. A Deep Reinforcement Learning (DRL) method was developed [4] to improve the accuracy of yield prediction. But the computational efficiency of the yield prediction was not improved. The Convolutional Neural Networks (CNNs) were developed [5] for the crop yield prediction. The designed model failed to train on a larger set of samples for achieving higher prediction accuracy. An artificial neural network (ANN) was developed [6] for effective agricultural yield prediction. But the performance of prediction time was not minimized. A Multi-tier machine learning architecture was designed [7] for crop yield prediction. But the deep feature learning was not performed to further improve the prediction accuracy. A random forest (RF) algorithm was designed [8] for cotton yield prediction at three distinct times. The designed algorithm minimizes the error rate of cotton yield prediction. However, it failed to use an effective tool for predicting the crop yield faster and precise manner. Coupling crop modeling and machine learning (ML) were developed [9] to enhance the corn yield predictions. But it failed to use the deep learning model for precise corn yield predictions.
Random forest methods were developed [10] to forecast the different crop yields such as wheat, barley, and canola. But the model failed to consider the huge time-series data for improving the quality of yield prediction.
The main contribution of the QRECF-DFFMPC are listed below
- •
To minimize the crop yield prediction time, the QRECF-DFFMPC technique uses the congruence correlative empirical orthogonal function for finding similar features. The correlation function is used to find the relevant or irrelevant features from the dataset. The relevant features are used for crop yield prediction instead of using entire features in the dataset.
- •
To improve the prediction accuracy and minimize the false positive rate, QRECF-DFFMPC uses the Feedforward multilayer perceptron to analyze the testing and training data at the hidden layer. The data points are analyzed using the Quantile regression function. Based on the regression analysis, the crop yield is correctly predicted.
- •
The performance of the proposed QRECF-DFFMPC technique is evaluated with different metrics and the results show that the proposed technique outperforms well than other baseline approaches.
- •
To overcome these issues of the existing literature a novel QRECF-DFFMPC technique was introduced in this work.
The paper is organized into five different sections as follows; Section 2 describes the related works in the field of crop yield prediction in precision agriculture. Section 3 discusses the proposed methodology QRECF-DFFMPC for predicting crop yield. In Section 4, experimental settings and data set descriptions are presented. Section 5 discusses the experimental results of the proposed methodology against the conventional methods. At last Section 6, the conclusion is presented.
Section snippets
Related works
A new multilevel deep learning approach integrated with RNN and CNN was introduced [11] for forecasting the corn yield based on spatial and temporal features. The designed approach reduces the mean square error. But the model failed to find the other crop yields such as soybean and winter wheat.
Process-based and remote sensing-driven crop yield models for maize (PRYM–Maize) were introduced [12] to evaluate the area maize yield prediction. But it failed to apply the machine or deep learning
Methodology
Agriculture plays a leading part in the growth of the country's financial system. In general, the agricultural yield mainly depends on weather conditions. Climate change has become the most important risk in agriculture. Therefore, crop yield prediction is the great importance to improve large-scale food production. An accurate prediction is to construct timely import and export decisions to improve national food security. However, several issues face the increased complexity of disease
Experimental setup
Experimental evaluations of the proposed QRECF-DFFMPC technique and two existing methods namely [1] and [2] are implemented using Java language. Java is extremely adaptable employed for programming applications on the web, mobile, desktop by using dissimilar platforms. In addition, Java includes several features such as dynamic coding, multiple security features, platform-independent characteristics, and so on. So, we have used Java language to implement the crop yield prediction with higher
Results and discussion
In this section, the experimental results of the proposed QRECF-DFFMPC technique and existing methods namely BBI-model [1], MANNs-SVR [2] are discussed in this section with different parameters such as prediction accuracy, error rate, and prediction time. Performance results of these metrics analyses are carried out with the help of tables and graphical representations. The analysis of three different methods is described in the following sections.
Conclusions
An accurate prediction of various specified crop yields helps to decide for long Term cultivation. To predict crop yields in the agriculture sector, a novel technique called QRECF-DFFMPC is developed. Deep FeedForward Multi-Layer Perceptron is used to deeply analyze the features in the hidden layer with the help of congruence correlative empirical orthogonal function. The correlative function finds the more relevant attributes and removes the other attributes from the dataset to minimize the
Authors biography
V. Sivanantham is currently a researcher scholar, working under the supervision of Dr. V. Sangeetha, Department of Computer Science, Periyar University Constituent College of Arts and Science, Pappireddipatti Campus, Periyar University, Tamil Nadu, India.
V. Sangeetha is currently working as a Assiatnt Professor in the Department of Computer Science, Periyar University Constituent College of Arts and Science, Pappireddipatti Campus, Periyar University, Tamil Nadu, India. She has more than 10
Declaration of Competing Interest
The authors declare that no conflict of interest.
Acknowledgments
The authors extend their appreciation to the Researchers supporting project number (RSP-2021/384) King Saud University, Riyadh, Saudi Arabia.
References (24)
- et al.
An end-to-end model for rice yield prediction using deep learning fusion
Comput Electron Agric
(2020) - et al.
Crop yield prediction using aggregated rainfall-based modular artificial neural networks and support vector regression
Environ Dev Sustain
(2020) - et al.
Machine learning for large-scale crop yield forecasting
Agric Syst
(2021) - et al.
Crop yield prediction using deep reinforcement learning model for sustainable agrarian applications
IEEE Access
(2020) - et al.
Crop yield prediction with deep convolutional neural networks
Comput Electron Agric
(2019) - et al.
Efficient agricultural yield prediction using metaheuristic optimized artificial neural network using Hadoop framework
Soft Comput
(2020) - et al.
Crop yield prediction: two-tiered machine learning model approach
Int J Inf Technol
(2019) - et al.
Crop yield prediction in cotton for regional level using random forest approach
Spatial Information Research
(2021) - et al.
Coupling machine learning and crop modeling improves crop yield prediction in the US Corn Belt
Sci Rep
(2021) - et al.
An approach to forecast grain crop yield using multi‑layered, multi‑farm data sets and machine learning
Precision Agriculture
(2019)
Multilevel deep learning network for county-level corn yield estimation in the U.S. corn belt
IEEE J Sel Top Appl Earth Obs Remote Sens
Developing a process-based and remote sensing driven crop yield model for maize (PRYM–Maize) and its validation over the Northeast China Plain
J Integr Agric
Cited by (8)
Data quality assessment and analysis for pest identification in smart agriculture
2022, Computers and Electrical EngineeringCitation Excerpt :Smart agriculture is facing new development opportunities in recent years [1–3].
An innovative smart agriculture system utilizing a deep neural network and embedded system to enhance crop yield
2024, Multidisciplinary Science Journal
- ☆
This paper is for special section VSI-hci2. Reviews were processed by Guest Editor Dr. Carlos Montenegro and recommended for publication.