Optimization of Depression Level Detection Based on Noninvasive Parameters with Artificial Neural Network
Depression among college students is increasing, which has an impact on decreasing productivity and even attempting suicide. It can reduce the quality and quantity of the productive age generation. Proper prevention and treatment of depression rely on establishing medical diagnoses that require detection instruments to maintain high accuracy. We presented a depression level measuring instrument that provides accurate decisions on student depression levels by implementing an Artificial Neural Network (ANN). Considering its use in students who require large-scale and high-frequency use, we have chosen non-invasive physical parameters. This study used an exploratory method to develop an ANN-based student depression level detector. The input layer of the ANN-based Depression Detector (dANN) received data on four non-invasive physical parameters, namely heart rate, sleeping hours, respiration rate, and blood oxygen, then the hidden layer ANN processed and learned deeply the data to present the results to the output layer which gives decisions in the form of 5 levels of depression. The learning results using Loss Graphs showed that dAAN reduced prediction errors or training iterations at Epoch greater than 130. dANN offers high accuracy, after reaching Epoch 130. Thus, a depression detection instrument has been successfully developed using ANN-based (dANN), prioritizing high accuracy in measurement practicality. dANN has the potential to be further developed by integrating psychological parameter data, considering that depression has a multifactorial character. Furthermore, dANN will be built on a web basis to improve performance by implementing an expert system.