Predicting Hand Grip Strength of Hand Held Grass Cutter Workers: Neural Network vs Regression

Exposure to hand transmitted vibration caused disability in term of hand grip strength force among hand held grass cutter workers. Objective: This current study develop prediction model of independent and dependent variable that induce to loss of grip strength using non-linear neural network and linear multiple regression prediction approach for both hands. Linear and non-linear approach was used the direct least square and activation sigmoid function, respectively. Method: 204 hand held grass cutter worker have been selected as the subject study due hand arm vibration exposure during operation which is significant to loss hand grip strength. The independent variables consist of age, height, weight, working experience and estimated vibration exposure per day while hand grip strength was selected as the dependent variables. Result: The performance indexes of regression are better fit for neural network compared to multiple regressions with 0.017 (right hand grip) and 0.066 (left hand grip) differences, respectively. The mean square error also stated near to “0” for non-linear compared to linear techniques. Conclusion: It concludes that the neural network model is superior to the linear model. However, best architecture of neural network algorithm could be implemented to increase performance index, hence produce the accurate prediction model for hand grip strength among grass cutter workers.