Using Data Mining to Predict the Concentration of Cadmium in Khuzestan Paddies, Iran

Abstract

Background: Rice is the second highly consumed foodstuff among Iranian people. However, high levels of cadmium (Cd) are reported in some paddy fields in Khuzestan province, Iran. Objectives: The current study aimed at investigating the Cd concentration in rice grains by the decision tree using J48 algorithm. The current study also used WEKA software to implement the algorithm. Methods: A total of 630 samples (9 attributes in 70 sampling areas) were taken from each paddy field (5 regions); hence, seed and soil samples were analyzed according to the standard laboratory procedures and finally, the data mining technique was used for the classification of trees by J48 algorithm to predict the concentration of Cd in rice seed. Results: The results showed that the average concentrations of Cd in rice seed and soil were 81.4 and 273.6 μg/kg, respectively; it was also shown that J48 gives 95.71% accuracy, 0.899 Kappa coefficient, and less error (RMSE = 0.179), which make a good predictive model. A significant correlation was observed between soil characteristics and the concentration of Cd in rice seeds. Conclusions: The data mining technology can be used to predicate Cd concentration in rice seeds, and also J48 algorithm is a simple designer to construct a decision tree; nevertheless, offers good results in experiments.