Application of unsupervised weighting algorithms for identifying important attributes and factors contributing to grain and biological yields of wheat

To identify important attributes/factors that contribute to grain and biological yields of wheat, 9912 sets of diverse data from field studies were extracted, and supervised attribute-weighting models were employed. Results showed that when biological yield was the output, grain yield, nitrogen applied, rainfall, irrigation regime, and organic content were the most important factors/attributes, highlighted by 9, 7, 5, 3 and 3 weighting models, respectively. In contrast, when grain yield was the output, biological yield, location, and genotype were identified by 8, 6, and 5 weighting models, respectively. Also, five other features (cropping system, organic content, 1000-grain weight, spike number m-2 and soil texture) were selected by three models as the most important factors/attributes. Field water status, such as the irrigation regime or the amount of rainfall, was another important factor related to the biological or grain yield of wheat (weight ≥ 0.5). Our results showed that attribute/factor classification by unsupervised attribute-weighting models can provide a comprehensive view of the important distinguishing attributes/factors that contribute to wheat grain or biological yield. This is the first report on identifying the most important factors/attributes contributing to wheat grain and biological yields-using attribute-weighting algorithms. This study opened a new horizon in wheat production using data mining.

Application of unsupervised weighting algorithms for identifying important attributes and factors contributing to grain and biological yields of wheat Keywords:

attribute weighting , Data Mining , unsupervised model , Wheat