Corn is an essential agricultural commodity since it is used in animal feed, biofuel, industrial processing, and the manufacture of non-food industrial commodities such as starch, acid, and alcohol. Early detection of diseases and pests of corn aims to reduce the possibility of crop failure and maintain the quality and quantity of crop yields. A decision tree is a nonparametric classification model in statistical machine learning that predicts target variables using tree-structured decisions. The performance of this model can increase significantly if the continuous predictor variables are discretized into valid categories. However, in some cases, the result does not provide satisfactory performance. The possible cause is the ambiguity in discretizing predictor variables. The incorporation of fuzzy membership functions into the model to resolve discretization ambiguity issues. This work aims to classify diseases and pests of corn plants using the decision tree model and improve the model’s performance by implementing fuzzy membership functions. The main contribution of this work is that we have shown a significant improvement in the decision tree model performance by implementing fuzzy membership functions; S-growth, triangle, and S-shrinkage curves. The proposed fuzzy model is better than the decision tree model, with an average performance increase from the largest to the smallest; kappa (12.16%), recall (11.8%), F-score (9.71%), precision (5.08%), accuracy (3.23%), specificity (1.94%), and AUC (0.49%). The combination of bias and variance generated by the proposed model is quite small, indicating that the model is able to capture data trends well.
Statistical machine learning has developed into integral components of contemporary scientific methodology. This integration provides automated procedures for predicting phenomena, case diagnosis, or object identification based on previous observations, uncovering patterns underlying data, and providing insights into the problem. Identification of corn plant diseases and pests using it has become popular recently. Corn (Zea mays L) is one of the essential carbohydrate-producing foodstuffs besides wheat and rice. Corn plants are sensitive to pests and diseases, resulting in a decrease in the quantity and quality of the production. Eradicate pests and diseases according to their type is a solution to overcome the problem of disease in corn plants. This research aims to identify corn plant diseases and pests based on the digital image using the Multinomial Naïve Bayes and K-Nearest Neighbor methods. The data used consisted of 761 digital images with six classes of corn plants disease and pest. The investigation shows that the K-Nearest Neighbor method has a better predictive performance than the Multinomial Naïve Bayes (MNB) method. The MNB method with two categories has an accuracy level of 92.72%, a precision level of 79.88%, a recall level of 79.24%, F1-score 78.17%, kappa 72.44%, and AUC 71.91%. Simultaneously, the K-Nearest Neighbor approach with k=3 has an accuracy of 99.54 %, a precision of 88.57%, recall 94.38%, F1-score 93.59%, kappa 94.30%, and AUC 95.45%.
Diabetes is a chronic disease that can cause serious illness. Women are four times more likely to develop heart problems caused by diabetes. Women are also more prone to experience complications due to diabetes, such as kidney problems, depression, and decreased vision quality. Nearly 200 million women worldwide are affected by diabetes, with two out of five affected by the disease being women of reproductive age. This paper aims to predict women with at least 21 years of age having diabetes based on eight diagnostic measurements using the statistical learning methods; Multinomial Naive Bayes, Fisher Discriminant Analysis, and Logistic Regression. Model validation is built based on dividing the data into training data and test data based on 5-fold cross-validation. The model validation performance shows that the Gaussian Naïve Bayes is the best method in predicting diabetes diagnosis. This paper’s contribution is that all performance measures of the Multinomial Naïve Bayes method have a value greater than 93 %. These results are beneficial in predicting diabetes status with the same explanatory variables.
In actuarial and insurance literatures, several researchers suggested generalized linear regression models (GLM) for modeling claim costs as a function of risk factors. The modeling of claim costs involving both zero and positive claims experience has been carried out by fitting the claim costs collectively using Tweedie model. However, the probability of zero claims in Tweedie model is not allowed to be fitted explicitly as a function of explanatory variables. The purpose of this article is to propose the application of Zero Adjusted Gamma (ZAGA) and Zero Adjusted Inverse Gaussian (ZAIG) regression models for modeling both zero and positive claim costs data. The models are fitted to the Malaysian motor insurance claims experiences which are divided into three types namely Third Party Bodily Injury (TPBI), Own Damage (OD) and Third Party Property Damage (TPPD). The fitted models show that both claim probability and claim cost are affected by either the same or different explanatory variables. The fitted models also allow the relative risk of each rating factor to be compared and the low or high risk vehicles to be identified, not only for the claim cost but also for the claim probability. The AIC and BIC indicate that ZAIG regression is the best model for modeling both positive and zero claim costs for all claim types
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.