Identifying healthy and mastitis Sahiwal cows using electro-chemical properties: A connectionist approach

This paper describes feed-forward sigmoid connectionist models to classify healthy and mastitis Sahiwal cows using pH, electrical conductivity, temperature (udder, milk and skin), milk somatic cells, milk yield and dielectric constant. Mastitis was determined according to two criteria: Somatic Cell Counts (SCC) over 2,00,000/ml and SCC over 5,00,000/ml. Cows with milk SCC below 2,00,000 were categorised as healthy cows while those with SCC ranging between 2,00,000 and 5,00,000 per ml were categorised as subclinical mastitis cows. The rest of the cows having milk SCC above 5,00,000/ml were considered under clinical mastitis category. The connectionist models were based on Error Back Propagation (EBP) learning algorithm with Bayesian Regularisation (BR) mechanism. Also, Multiple Linear Regression (MLR) models were developed to classify the healthy and mastitis Sahiwal cows for comparing the classification accuracy of proposed connectionist models. As a result, the connectionist approach was found to be more effective than the conventional regression technique for classifying healthy and mastitis Sahiwal cows.