Network is fully loaded with traffic which needs to be classified and it has become the most critical aspect in improving the security, performance, and resource management requirements within a modern communication network. This paper presents a new approach to the application-based network traffic classification that integrates advanced preprocessing techniques with the use of Ant-Lion Optimization (ALO) for feature extraction and fully connected encoders for classification. The preprocessing step includes data cleaning, normalization, and dimensionality reduction which ensure efficient handling of high-dimensional traffic data. ALO helps optimize feature selection to better identify key attributes that can distinguish different application classes. The learning robust feature representation and accurate classification of different classes of network traffic with a Fully Connected Encoder (FCE) architecture is followed. Experimentation show the approach significantly outperform traditional methods in terms of classification accuracy, computational efficiency, and scalability. This method is expected to accurately classify network traffic in real time and exhibits robust performance even in dynamic environments of networks.