Developers can use software defect prediction, which predicts problematic code sections, to detect flaws and prioritize their testing efforts. Traditional fault prediction algorithms frequently fail to account for semantic changes across programmes. The traditional model of prediction fails to adapt newer bugs . To create accurate prediction models, the ability to capture semantics in programmes is essential. In order to predict the new age we suggest using a deep learning mechanism which can properly give weight age to the semantic features and understand them in detail so that newer combating technologies can be developed We use a deep Keras Library to learn semantic features automatically using token vectors taken from the text file we provided. The model is then properly predicted using DBN and classifiers.