MOHAN KUMAR, T. L.KARTHIK, M. N.2023-01-142023-01-142022-01-13https://krishikosh.egranth.ac.in/handle/1/5810192300The present study was undertaken to analyse the distribution pattern, correlation with weather parameters and prediction of fruit fly incidence on bitter gourd during Kharif season for the data collected from IIHR, Hessaraghatta, Bangalore for the period of three year (2018-2020). The average, maximum and minimum incidences of fruit fly was observed 47.81, 138 and 2 number respectively over the study period. One way ANOVA revealed that the average fruit fly incidence was found to be non-significant among the three study years. To determine the distribution pattern of fruit fly incidence, discrete probability distributions viz. Poisson, geometric, negative binomial, logarithmic, zeta and Yule-Simon distributions were employed. The negative binomial distribution was found to be the best fitted distribution during 2018 and 2019, whereas geometric distribution was best fitted distribution during 2020. Correlation analysis revealed that the fruit fly incidence across the year showed positive correlation with maximum temperature, morning relative humidity, evening relative humidity and evaporation whereas, minimum temperature, wind speed and rainfall showed negative correlation. Multiple Linear Regression (MLR), Support Vector Regression (SVR) and Artificial Neural Network (ANN) models were employed to predict fruit fly incidence based on weather parameters. Based on the lowest MAPE value, ANN model was found to be best-fitted model followed by SVR and MLR models on both training (16.82 %) and testing datasets (6.33 %), which indicates superiority of ANN model over other models. Therefore, ANN model can be used for predicting fruit fly incidence in bitter gourd.EnglishThesis