The Parameterization of a Model for Wild Chickpea Flowering Time by Transferring Knowledge from Multiple Sources

Abstract

Forecasting flowering time allows researchers to create plant varieties that achieve maximum efficiency and value in the face of climate change. In this paper, we propose an algorithm for parameterizing the flowering time model of wild chickpea samples, which uses the transfer learning technique to combine several sets of source data and target data. The constructed model, using genetic and climatic data only for the first 20 days after sowing, predicts the flowering time of the samples with high accuracy; the average absolute error is slightly more than 5 days and the Pearson correlation coefficient is 0.93. It was found that the maximum and minimum temperatures have the strongest effect on the flowering time. At the same time, all weather factors on the seventh–tenth day after sowing influence the solution of the model.