Paper Title
APPLICATION OF NON-LINEAR MACHINE LEARNING STATISTICAL MODELS IN HORTICULTURAL CROP RESEARCH

Abstract
Horticulture sector contributes to 33% of Indian agriculture GDP, as the total Indian horticulture crop production itself is projected to increase from the present 370.73 mt to 777.77 mt by 2047. Being a perishable product with enormous variability among its genetic resource, its research and development sector depends heavily on high end & efficient statistical techniques to exploit this variability resulting in new improved technologies and varieties ultimately to increase the productivity from the present 12.3 to 18.4 by 2047. More specifically, statistical methods play a vital role in different areas of horticultural crops research right from planning of any research experiment systematically till selection of appropriate tool for data analysis to unravel many of the hidden results.They are central to horticultural research, enabling robust analysis of complex research issues entailing multi-disciplinary fields such as crop improvement, production, protection, post-harvest management, climate resilience, and socio-economic studies.Further, statistical models are backbone for assessing the influence of set of traits on increasing any crop productivity and to arrive at the significant traits with their optimum values, a backbone for precision horticulture. As the inter-relationship among the traits, within and across crop phenological stages/seasons/years, are non-linearly related, machine learning based statistical models captures the reality in superior way and subsequently increase the prediction power. In this direction, attempts were made and in progress at ICAR-IIHR, Bengaluru to develop suitable ML/ANN based non-linear statistical models in various areas of horticultural research like canopy modeling, crop-logging studies, disease/pest forecasting studies, non-destructive leaf area prediction in perennials, carbon sequestration studies, non-parametric stability analysis, construction of selection indices, genomic selection models, mass transfer kinetics etc. The present communication, while delineating several of such multidisciplinary research outcomes arrived at ICAR-IIHR, Bengaluru, also provides a road map for its potential application in emerging areas of research to achieve sustainable crop productivity. Keywords - Carbon Sequestration, Crop modelling, Crop-varietal release, Horticulture, Machine learning models, Mass Transfer Kinetics, Statistical models