A comparison of artificial neural networks learning algorithms in predicting tendency for suicide

New approaches adopted by behavioral science researchers to use modern modeling and predicting tools such as artificial neural networks have necessitated the study and comparison of the efficiency of different learning algorithms of these networks for various applications. By using well-known and different learning algorithms, this study examines and compares the Perceptron artificial neural network as predicting tendency for suicide based on risk factors within 33 input parameters framework used in neural network. To find the "best" learning algorithm, the algorithms were compared in terms of train and capability. The experimental data were collected through questionnaires distributed among 800 university students. All questionnaires used in this research were standardized with appropriate validity and reliability. The study findings indicated that LM and BFG algorithms had close evaluation in terms of performance index and true acceptance rate (TAR), and they showed higher predictive accuracy than the other algorithms. Furthermore, CFG algorithm had the minimum training time.