Dr. A. Mahendar
Department of CSE (Data Science), CMR Technical Campus, Kandlakoya, Hyderabad, Telangana 501401, India

Published : 1 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 1 Documents
Search

Intelligent Crop Yield Level Prediction Using a Zebra Optimization-Based Multilayer Perceptron Prediction Mechanism (ZMPPM) K.Venkata Rama Rao; Dr Lakshmi Devi N; Dr. A. Mahendar; Chavvakula Janaki Devi; Jagan Mohan Sarihaddu; Dr. V. Gokula Krishnan
Asian Journal Science and Engineering Vol. 5 No. 2 (2026): Asian Journal Science and Engineering
Publisher : CV. Creative Tugu Pena

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.51278/ajse.vi.2307

Abstract

Accurate yield prediction has a significant impact on the overall improvement of agricultural planning better resource management, and even food security. Here a new Zebra Multilayer Perceptron Prediction Mechanism (ZMPPM) is presented as a method of realizing effective yield prediction from historical and environmental data. The framework comprises four stages: data preprocessing, Zebra Optimization Algorithm (ZOA)-based feature selection, ZOA-assisted training of a multilayer perceptron (MLP), and classification of crop yield into Low and High yield levels. In ZMPPM, the ZOA is used (i) to select the most informative subset of soil, climatic and management variables and (ii) to optimise the learning parameters of the MLP, which then captures the nonlinear relationships between these variables and yield. The model was evaluated on a historical crop yield dataset of soil, climatic and management variables using a stratified hold-out test set of 5,000 records. On the test set, ZMPPM achieved an accuracy of 99.07%, a precision of 99.40%, a recall of 99.30% and an F1-score of 99.35%. Compared with Decision Tree, Gradient Boosting Machine, XGBoost, LSTM and CNN-LSTM baselines, ZMPPM improved accuracy by 1.50 percentage points over the best baseline (LSTM, 97.57%). These results indicate that ZMPPM is a promising component for agricultural decision-support systems, although validation on additional regions and seasons is still required.