Nowadays, software defects are considered the major problem in software that leads to system failures. The prediction of software defects is becoming more popular due to its significant role in increasing the reliability and quality of software. Nowadays, various studies are introduced to support software developers in predicting software defects. Meanwhile, these techniques ignore the semantic as well as contextual information of source code during Software Defect Prediction (SDP). To address this issue, an optimization-enabled Machine Learning (ML) model Adam Northern Goshawk Optimization-enabled Rider Neural Network (ANGO-RideNN) is designed in this research to accurately predict software defects. Here, the input software data considered from the dataset is subjected to the selection of suitable features using the Relief-wrapper (RW)-based feature selection technique. From the selected features, the software defects are effectively detected using Rider Neural Network (RideNN). Moreover, the prediction performance of RideNN is increased by training the weights using Adam Northern Goshawk Optimization (ANGO). Further, the prediction performance of ANGO-RideNN is validated and the experimental results proved that the ANGO-RideNN attained high prediction results with Mean Magnitude of Relative Error (MMRE), Mean Absolute Percentage Error (MAPE), and Root Mean Squared Error (RMSE) of 0.102, 0.142, and 0.224, respectively.