In-materio reservoir computing (IMRC) is a promising system of low-power learning method. Here, we investigated the relationship between light irradiation and task accuracy using an oriented thin film of the photoresponsive material Poly(3-hexylthiophene-2,5-diyl) (P3HT) as the reservoir layer. It was found that P3HT can represent a more complex waveform under light irradiation. In the NARMA task of time series prediction, the waveforms were sufficiently predicted, and the accuracy was improved by light irradiation.