A significant influence on recommendation systems comes from the attractiveness of products in the current rating collection (RS). To lessen the recommendation model’s bias, it is also necessary to address the diversity approach, which asks for adding less popular items to the list of ranked items during recommendation. Unlike traditional RS, the Recommender System filter’s range of aspect-based enhancement may alter the user’s dispersion ratio. This study presents a probabilistic approach to predict users’ preferences when selecting carefully ranked items. In the suggested procedure, a list of 10 things is evaluated in order of popularity. In order of decreasing popularity, the first seven items are chosen for recommendations. The item’s popularity and likelihood are calculated using the Hidden Markov Model (HMM). The last three options on the recommendation list are chosen from the less likely list, which is generated as well. The phases in the suggested technique are user interest pattern modelling, missing filling entry, and likelihood estimate using an HMM model. The findings of a comparison study using a graphical representation of user dispersion for the EDUA and CML parameters show that there are more users for less likely items when compared to modern methods.