A company is using a cloud-based big data platform to analyse user behaviour, e.g., browsing history, clicks, and purchase records. The system processes large-scale datasets to generate personalised product recommendations.

Which approach do you think is the most commonly used to build such recommendation systems in the cloud? (Select TWO)

Question

A company is using a cloud-based big data platform to analyse user behaviour, e.g., browsing history, clicks, and purchase records. The system processes large-scale datasets to generate personalised product recommendations.

Which approach do you think is the most commonly used to build such recommendation systems in the cloud? (Select TWO)

BlackTom AI · Accepted Answer

Scalable analytics pipelines plus ML models are the standard design pattern for cloud-based recommendation systems, so the correct selections are the first and fourth options.

1. Relevant knowledge points and concepts
- Big data analytics: ingesting, storing, and processing massive behavioural logs (clickstreams, browsing, purchases) using distributed storage (S3/HDFS) and computation (Spark, Flink).
- Machine learning for recommendations: collaborative filtering, matrix factorization, content-based models, and hybrid models that learn user/item embeddings and predict preferences.
- Cloud advantages: elastic compute, managed data pipelines, feature stores, model training and serving (batch and real-time).

2. Step-by-step reasoning
- Step 1: Collect and store large-scale user-event data (events per second, TBs/day) — requires big-data tools, not per-user manual handling.
- Step 2: Preprocess and aggregate data (sessionization, feature extraction) using distributed analytics (Spark, Beam).
- Step 3: Train ML models (e.g., ALS, neural recommenders) on aggregated features to learn user/item patterns.
- Step 4: Deploy models to serve recommendations in real time or generate batch recommendations periodically.

3. Intermediate steps / concrete example
- Example pipeline: raw events → feature extraction (user click counts, item co-occurrence) → train collaborative filtering model → compute top-N recommendations → serve via API.
- Scalability: processing 100M events requires parallel computation; manual per-user rules or single relational DB queries would bottleneck.

4. Why the correct answers are correct
- "Using big data analytics…" is necessary because the datasets are large-scale and require distributed processing.
- "Applying machine learning algorithms…" is required to identify latent patterns and produce personalized ranking; ML generalizes across users and items efficiently.

5. Why other options are incorrect
- "Manually analysing each user’s data…" is infeasible at scale and cannot adapt automatically.
- "Storing user data in relational databases and manually querying it…" is too slow and unscalable for high-throughput recommendation workloads; relational queries alone don’t provide predictive ranking.

6. Conclusion
The most commonly used approaches are big data analytics and machine learning, matching the provided answers: "Using big data analytics to process large-scale user behaviour data" and "Applying machine learning algorithms to identify user patterns and preferences."

类似问题

When a new video is posted on TikTok, how does the platform's algorithm typically begin distributing that content?

What is the primary factor that drives the TikTok algorithm when deciding which videos appear on a user’s For You Page?

What factors affect what users see in their Instagram news feed the most?

When social media algorithms recommend news, what are their recommendations based on?

TikTok’s algorithm considers watch time on videos to decide recommendations.

推荐系统与传统搜索引擎有哪些不同点？（多选题） What are the differences between recommender systems and traditional search engines? Select all that apply

What is a Recommendation Engine used for?

Question at position 10 Vicky explains that everybody's content is different because algorithms identify people's Question Blank 1 of 1[input] from their personal data.

更多留学生实用工具

智能学习助手

风格化写作助手

论文查重助手

文献引用助手

课堂转译助手

课堂笔记助手

Quiz搜索助手

学校历年真题

智能刷题助手

智能匹配练习

希望你的学习变得更简单