Online Bangla news has rapidly increased in the era of the information age. Each news site has its different categorization for grouping their news. The Layout and categorization of the online Bangla news articles cannot perpetually meet the individual's needs due to the heterogeneity. So, overcoming this issue and classifying the online Bangla news articles according to the preference of the user is an arduous task. So, it is essential to provide state-of-the-art solutions as well as the best way to solve this problem. The paper aims to build an automated system to classify the Bangla news contents and also find out the state-of-the-art solutions for the small size dataset. It is known that most of the machine learning models need huge amounts of data for the proper training and testing of the models. But due to the scarcity of the dataset, it is not always possible to provide the state-of-the-art solutions. But, in this research work, we have found that Text-GCN performed better than the BiLSTM, GRU-LSTM, LSTM, Char-CNN, and BERT to classify the online Bangla news in spite of the small size of the dataset. The obtained experimental result shows the efficiency of the Text-GCN over the other models in terms of accuracy, precision recall, and F1-score.