Prediction of Raw Milk Microbial Quality Using Data Mining Techniques

Abstract

Data mining techniques were applied to predict raw milk quality interms of methylene blue reduction time (MBRT) from the independentparameters of raw milk inspection parameters such as travel time,temperature of milk, solid-not-fat, %fat, acidity and specificgravity. Predictive models were developed and the performance of 3 datamining algorithms namely; Multiple Linear Regression (MLR), ArtificialNeural Network (ANN) and K-Nearest neighbor (KNN), was measured interms of average error and Root Mean Square Error (RMSE). MLR showedhigh and inconsistent RMS error in 3 randomly picked data partitionswhereas KNN and ANN were able to predict the MBRT values from thephysico-chemical quality parameters, KNN was the preferred algorithm(K=7, RMSE of 1.7). The models were applied to a new set of data (n=78)without showing them the output parameter (MBRT). The predicted valuesof MBRT were plotted against the actual observed values to classifymilk into 4 quality grades.