--- title: "Generic to personalized recommender" subject: "Information System" chapter: 7 key: "semester_8/information_system:recommender-generic-to-personalized" source: https://www.bctengineer.com/q/semester_8/information_system/recommender-generic-to-personalized site: BCT Engineering --- # Generic to personalized recommender Information System · Chapter 7: Web based information system and navigation - Asked in: 2073 Magh, 5+5 marks - Video: [Recommendation System Machine Learning|Recommendation Engine basics In Machine Learning](https://www.youtube.com/watch?v=juqpTaieIkA) — Unfold Data Science - Answer checked by Claude Opus 5.5 (high effort) on 2026-10-14 - Page: https://www.bctengineer.com/q/semester_8/information_system/recommender-generic-to-personalized - Whole chapter: https://www.bctengineer.com/md/semester_8/information_system/chapter-7 ## Question What is recommender system? What specifics are required to transform a generic recommender system into a personalized recommender system? Illustrate with specific examples. ## Short answer A recommender system suggests items to users. To personalize, add user identification, profile building from explicit and implicit data, item profiles, personalization algorithms (content-based, collaborative), context, feedback loops, and cold-start handling. Example: instead of generic 'Top 10' news, show sports to Ram and politics to Sita based on past reading. ## Answer #### Recommender system A **recommender system** is a software tool that suggests items (products, movies, books, news) likely to interest a user, to help the user cope with too many choices. It works from user preferences, item features and the behaviour of other users. A **generic (non-personalised)** recommender gives the same recommendations to everybody, such as "top 10 best sellers", "most viewed" or "new arrivals". A **personalised** recommender gives each user a different list based on his or her own profile and behaviour. #### What is required to transform a generic recommender into a personalised one 1. **User identification:** login, cookies or device ID so that each user's activity is recorded. 2. **User profile building:** store explicit data (rating, interests, demographics) and implicit data (clicks, purchases, watch time, search history) for each user. 3. **Item profiles:** features such as genre, price, brand, keywords, tags. 4. **Personalisation algorithms:** content-based filtering, collaborative filtering or hybrid, in place of simple popularity counts. 5. **Context information:** time, location, device and mood. 6. **Feedback loop:** continuous learning from new actions (likes, skips, purchases). 7. **Handling cold start:** use onboarding questions, demographics, or popularity for new users. 8. **Scalable infrastructure:** real-time computation, caching and storage for each user. 9. **Privacy and consent controls,** and explanation of the recommendations. 10. **Evaluation:** A/B testing, click-through rate, conversion. #### Specific examples - **Generic:** a news website showing "Top 10 stories today" to everyone. - **Personalised:** the same website showing Ram sports and Sita politics, based on past reading. - **E-commerce:** a generic list shows best-selling phones. After the user views camera phones under Rs. 30,000, the personalised system shows similar camera phones in that range, and accessories bought by similar users. - **Music:** instead of "Top 50", Spotify builds a playlist from the user's listening history and similar listeners. ## Check yourself The answer's key points as multiple-choice questions, with the correct option and why. 1. What is a generic recommender? A. One that gives a different list per user B. One that gives the same recommendations to everybody, e.g. top 10 best sellers C. One that needs login D. One that uses watch history Answer: B. A personalised one uses each user's profile and behaviour. (Key point: Generic vs personalised) 2. What is needed first to personalise? A. User identification (login, cookies, device ID) to record each user's activity B. A larger screen C. Deleting item data D. A top-10 list Answer: A. Activity must be linked to individual users. (Key point: First requirement) 3. Which algorithms replace simple popularity counts? A. Random selection B. Alphabetical sorting C. Date sorting D. Content-based, collaborative or hybrid filtering Answer: D. They produce per-user recommendations. (Key point: Algorithms) 4. Why include a feedback loop? A. To stop collecting data B. To make lists identical for all C. To learn continuously from new actions such as likes, skips and purchases D. To remove user profiles Answer: C. Learning from actions keeps recommendations current. (Key point: Feedback loop) ## Related in Information System - [Recommender system design and types](https://www.bctengineer.com/question.md?key=semester_8%2Finformation_system%3Arecommender-design-types) — Information System, chapter 7 - [Recommender systems and examples](https://www.bctengineer.com/question.md?key=semester_8%2Finformation_system%3Arecommender-how-works-examples) — Information System, chapter 7 ## Related in other subjects - [ILD as system-specific; focal point recurrence](https://www.bctengineer.com/question.md?key=semester_5%2Ftheory_of_structures_ii%3Aild-system-specific-focal-point-recurrence) — Theory of Structures II, chapter 4 - [Hump check using given specific energy curve](https://www.bctengineer.com/question.md?key=semester_4%2Fhydraulics%3Ahump-adequate-specific-energy-curve-given) — Hydraulics, chapter 7 - [Block diagram of digital communication system](https://www.bctengineer.com/question.md?key=semester_5%2Fdata_communication%3Ablock-diagram-digital-comm) — Data Communication, chapter 1, asked 8× - [Inverse Z-transform by inversion integral](https://www.bctengineer.com/question.md?key=semester_6%2Fdigital_control_system%3Ainverse-integral-step-exponential) — Digital Control System, chapter 2, asked 10×