Automated health care products classification is a challenging field of research that has recently gained a lot of interest. This process is beneficial in terms of time and cost but problematic in obtaining annotated data and lacks uniformity. The high volume of healthcare products and their categories raise the need for machine learning models that can decrease the time and cost spent by human editors. Deep learning techniques that have recently emerged are applied to automated healthcare data classification. The efficacy of the deep learning model depends on the training data and the learning model's suitability for the data domain. When the dataset is large, training a model requires potent processors, including GPUs and might take hours. However, when such a large volume of data is unavailable, the Conventional Neural Network (CNN) does not train well for the lack of enough samples. To overcome this issue, an effective classification method is proposed to classify the products with a contemporary architecture that integrates the data selection, transformation, and filtering processes with the training of CNN and long short-term memory (LSTM) with limited labeled data and has an imbalance among the classes. Finally, the proposed approach demonstrates that using an efficient classifier at the end of the CNN structure delivers the desired performance even when the CNN model is not intensely trained.