Drug-induced peripheral neuropathy occurs as an adverse reaction of chemotherapy. However, a highly accurate method for assessing peripheral neuropathy and pain caused by compounds has not been established. The microelectrode array (MEA) assay using human induced pluripotent stem cell (iPSC)-derived neurons is expected to be one of the in vitro assays for predicting the mechanism of action (MoA) of pharmaceuticals. However, it is difficult to detect the reactions of drugs with different modes of action with a single parameter, and the analysis method is a challenge. To address this issue, acquiring more detailed information on neuronal network activity focused on individual cells is considered effective. In this study, human iPSC-derived sensory neurons and rat DRG neurons were cultured on 236,880 electrode CMOS-MEA, and the precise electrical activity of individual neurons was acquired. The relationship between the spontaneous activity pattern of peripheral neurons and the induced response pattern to agonists of TRPV1, TRPA1, and TRPM8, respectively, was analyzed. As a result, differences in agonist types were visualized by UMAP analysis based on spike patterns. Next, we developed a pain prediction AI that learned UMAP coordinate information and agonist type. The developed model predicted agonist type and concentrations with 81% accuracy. The CMOS-MEA, which can acquire precise electrical activity of single neurons, can increase the parameters that can detect the effects of drugs, making it effective as a method for predicting the MoA of compounds.