Answer-first summary for fast verification
Answer: Cosine Similarity
Cosine Similarity is the most suitable technique for this scenario because it measures the cosine of the angle between two vectors in a multi-dimensional space, effectively identifying how similar two products are based on their features or customer interactions. This method is particularly useful with limited data as it doesn't require a large dataset to produce meaningful similarity measures. - **A (PCA)** is primarily used for dimensionality reduction, not for measuring similarities between items. - **B (Matrix Factorization)** is effective for recommendation systems but typically requires more data to uncover latent factors. - **D (One-hot encoding)** is a preprocessing step for categorical data, not a method for calculating similarities. For further reading, consider exploring: - [Cosine Similarity in Recommender Systems](https://en.wikipedia.org/wiki/Cosine_similarity) - [Machine Learning Recommendations on Google Cloud](https://cloud.google.com/architecture/recommendations-using-machine-learning-on-compute-engine)
Author: LeetQuiz Editorial Team
Ultimate access to all questions.
As a Data Scientist at a retail company, you are tasked with enhancing the customer shopping experience by recommending products based on their purchasing history and preferences. The company has a vast inventory but limited historical data on customer purchases. You need a method that can efficiently identify product affinities to recommend complementary products, even with sparse data. Which of the following techniques would be most effective for this purpose? (Choose one correct option)
A
Principal Component Analysis (PCA)
B
Matrix Factorization
C
Cosine Similarity
D
One-hot encoding
No comments yet.