Document Detection using Probabilistic Principal Component Analysis-Based Naïve Bayes Model

Document detection is where algorithms are trained to detect and classify document types such that automating sorting and data extraction becomes more efficient in document management. Common limitations of existing methods include reliance on large volumes of labelled training data, sensitivity to differences in document formats, and problems with low-quality images. Therefore, this research performs the document detection using machine learning methodology. Initially, the data is manually collected from online sources in the form of documents together with their categories. These collected data are subjected to the pre-processing using the log-linear lemmatization approach. The features are extracted from these pre-processed data using the probabilistic Principal Component Analysis (PCA) approach. Finally, the document detection is accomplished using the Naive Bayes (NB) model that produces the final outcome. Simulation findings revealed that the proposed Probabilistic PCA-NB model returns better results than the other considered conventional methods respectively. The proposed Probabilistic PCA-NB for the document detection model is 0.62%, 0.72%, 0.71%, and 1.59% better than the other considered traditional methods in terms of accuracy, precision, sensitivity, and specificity respectively.