Machine learning model for estimating agricultural crop insurance payout based on air temperature, rainfall, and relative humidity

K.P. Mangani1*, R. Kousalya2

1Department of Computer Science, Dr.N.G.P Arts and Science College, Coimbatore, Tamilnadu, India
2Research Supervisor, Head of the Department of Computer Application, Dr.N.G.P Arts and Science College, Coimbatore, Tamilnadu, India

(Manuscript received 17 March 2020; accepted for publication 7 May 2020)

Abstract. In Agriculture, the weather-based variations are deliberated to estimate the crop insurance payout. This research model includes linear regression technique (LR) for air temperature payout prediction and fuzzy based choquistic regression (FCR) technique for rainfall payout prediction of agricultural blocks. Then the combined indices of rainfall, relative humidity and air temperature are considered as input to the proposed model named fuzzy based Quasi Poisson Regression technique (FQPR) implementing the multi-indices evaluation function that performs the total payout prediction per hectare of the specified block. The deviations in weather indices determine the insurance payout value with the threshold parameter specified as per policy makers. Thus, the proposed techniques can support the prediction of the total insurance payout with additional weather parameters for the seasonal period of the selected crop for selected five districts with reduced error rate. The results show that the proposed work is appropriate for combining weather indices and predicting the total insurance payout of the groundnut crop of the selected districts.

Plant cell walls fiber component analysis and digestibility of birdsfoot trefoil (Lotus corniculatus L.) in the vegetation

Y. Naydenova, A. Kyuchukova, D. Pavlov
Abstract. The changes in plant cell wall fiber components content and digestibility by Van Soest detergent analysis and in vitro enzymatic digestibility of Bulgarian plant breeding materials of birdsfoot trefoil (Lotus corniculatus L.) in the vegetation with a view to characterize plant species and to develop predictive regression models for forage quality evaluation are presented. The study was carried out at the Institute of Forage Crops, Pleven, as a part of its breeding program. The plants were grown during the period 2002–2004 on experimental plots and harvested at eight development stages from pasture stage to full pod formation stage in the first spring and second summer growths. During the vegetation both in spring and summer growths cell wall fiber components changes were presented as NDF, ADF, ADL. The rate in all parameters content increasing of the cell wall fiber components is great till full flowering stage, after that the intensity decreases. The degree of plant lignification is evaluated. Digestibility at each plant development stage in spring growth compared to those of summer is higher from 2,4 to 5,7 % points at average-day decreasing by 0,22% units (pasture stage-full flowering stage) in spring and 0,26% units (pasture stagebeginning of pod formation stage) in summer growth. The linear regression mathematical models predict in vitro dry matter digestibility by different cell wall fiber fractions with high accuracy (coefficient of determination R=0,847–0,937). When digestibility was predicted by all fractions, the predictive accuracy was the highest (R=0,986–0,994). Accuracy of estimation of fiber components and digestibility by days of vegetation as an independent variable was also high in spring growth R=0,859–0,944 and in summer growth R=0,906–0,989.