A robust E learning recommendation system based on novel interval valued bipolar fuzzy hypersoft set theory

Why smarter learning suggestions matter

Online courses, videos, and apps now flood the internet, but most learning platforms still struggle to suggest the right material to the right person at the right time. Many recommendation engines only see what you clicked “like” on, and ignore what you quietly disliked or felt unsure about. This paper presents a new mathematical framework that helps e‑learning systems better capture both sides of a learner’s feelings—what they welcome and what they want to avoid—while also dealing with uncertainty and the many layers of detail in course content. The goal is simple: make personalized learning suggestions that feel as if a thoughtful tutor knows you well.

Seeing both likes and dislikes

Most current recommendation methods focus on positive signals, such as ratings, completed lessons, or time spent on a page. But real learners also have dislikes and hesitations: a student might enjoy short interactive videos yet strongly avoid long text lectures, or like the pace of a course but be unhappy with its shallow quizzes. The authors argue that any system which ignores these negative or mixed reactions risks giving suggestions that feel tone‑deaf. Their work builds on a line of “soft” and “fuzzy” mathematical tools designed for messy, imprecise information. In this new approach, every learning option can carry both a degree of attraction and a degree of repulsion, each described not by a single number but by a range, to reflect uncertainty or disagreement among users.

Breaking content into fine-grained pieces

Learning materials are rarely simple. A course can be judged by its content, interactivity, feedback, and evaluation style, and each of these can be broken down further. For example, content has relevance, clarity, and engagement; feedback has specificity and frequency; evaluation can emphasize learning outcomes or satisfaction. The proposed framework, called an interval‑valued bipolar fuzzy hypersoft set, is built to mirror this layered structure. It lets the system handle combinations of sub‑features—such as highly engaging but only moderately clear videos with rich, frequent feedback—and keep track of how much each of these aspects is liked or disliked, again using ranges rather than precise scores. This creates a rich, structured picture of both learners and learning resources.

How the new scoring method works

On top of this structure, the authors design a decision algorithm that assigns an overall score to each learning option. First, the system records interval ranges for how positively and negatively each alternative is perceived along all relevant sub‑attributes. It then computes how wide these ranges are, capturing how certain or uncertain the judgments are. By summing and comparing interval lengths across rows and columns in special tables, the algorithm produces a positive score and a negative score for each option. Subtracting the negative from the positive yields a final value that can be used to rank choices. In a worked example, four instructional approaches—traditional classroom teaching, blended learning, self‑directed learning, and informal learning—are evaluated across multiple aspects such as content quality, interactivity level, and feedback type. The method identifies blended learning as the best overall alternative under the chosen criteria.

Putting the model to the test

To check whether this intricate structure is not just elegant but also sound, the paper examines its basic logical properties. The authors show that unions, intersections, complements, and combined "AND" and "OR" operations behave in predictable ways, satisfying commutative, associative, distributive, and De Morgan‑type laws. They then compare their recommendation approach with standard collaborative filtering and a popular fuzzy ranking method known as Fuzzy TOPSIS. Using synthetic data that imitate student preferences, they evaluate precision, recall, and F1 scores—common measures of how many good suggestions are found and how many bad ones are avoided. The new method achieves higher accuracy and broader coverage, largely because it can incorporate both positive and negative evidence along with uncertainty in a single coherent structure.

What this means for future e-learning

In plain terms, the study shows that carefully modeling what learners like, what they dislike, and how unsure they may be leads to smarter, more trustworthy recommendations. By representing these feelings as ranges instead of fixed scores, and by respecting the many layers of course design, the interval‑valued bipolar fuzzy hypersoft framework gives platforms a richer language for understanding learners. While the current results are based on controlled, synthetic data and rely on expert‑chosen ranges, the authors envision future systems in which real user data and machine learning help estimate these intervals automatically. If realized, such systems could power the next generation of e‑learning platforms, where recommendations feel less like generic lists and more like thoughtful guidance from a well‑informed mentor.

Citation: Harl, M.I., Saeed, M., Saeed, M.H. et al. A robust E learning recommendation system based on novel interval valued bipolar fuzzy hypersoft set theory. Sci Rep 16, 13239 (2026). https://doi.org/10.1038/s41598-026-42231-6

Keywords: e-learning recommendation, personalized education, fuzzy decision-making, learner preferences, multi-criteria ranking