Spike sorting is challenging due to the complexity of neural data requiring advanced algorithms in order to accurately separate and classify individual spikes from noisy recordings. Overlapping and invalid clusters are one of the hardest to solve, along with the similarity of spike shapes, artefacts, and imbalanced data. This paper presents a novel deep neural network approach to address these challenges and improve the overall performance of spike sorting. The proposed method can accurately classifying spikes, and it effectively deals with the classification problem of misclassified spikes and/or invalid clusters. The proposed method achieves a classification accuracy of 99.85%, outperforming other spike sorting methods.