Evaluation of Item-Based Top-N Recommendation Algorithms

By George Karypis. 2001

770 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Information Systems, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Information Systems

In Signal Processing

Citations in the corpus

Collaborative Topic Regression with Social Matrix Factorization for Recommendation Systems (2012, methodology): “The memory-based approaches use either user-based approaches (Herlocker et al., 1999) or item-based approaches (Karypis, 2001) for prediction (recommendation) of ratings for items.”
Explainable Recommendation: A Survey and New Perspectives (2018, background): “Readers can refer to Shani and Gunawardana [2011] and Karypis [2001] for a comprehensive survey of offline evaluation methods for recommendation systems.”
Data Mining for Web Personalization (2007, methodology): “Two commonly used metrics in this context are Precision and Recall which are standard metrics used in evaluating information retrieval effectiveness, but have also been adopted to evaluate ranked ordering of recommended items in personalization [58, 116, 15].”
The MovieLens Datasets: History and Context (2016, background): “…to develop and test many of the core algorithmic advances in recommender systems, including item item collaborative .ltering [Sarwar et al. 2001; Karypis 2001; Deshpande and Karypis 2004], dimensionality reduction collaborative .ltering [Sarwar et al. 2000], trust-based recommenders [O…”
An experimental investigation of kernels on graphs for collaborative recommendation and semisupervised classification (2012, methodology): “This method corresponds to the methodology proposed by the SUGGEST approach (Karypis, 2001).”
Research-paper recommender systems: a literature survey (2015, background): “Naturally, recommendation approaches perform differently on different datasets [224,292,293].”
Music recommendation and discovery in the long tail (2008, methodology): “Average Reciprocal Hit–Rate (ARHR) was first used in (Karypis, 2001).”
Visually Explainable Recommendation (2018, methodology): “We use the following measures for evaluation on different tasks: • Precision (P), Recall (R) and F 1 -score ( F 1 ): These measures aim to evaluate the quality of the recommendations [15].”
Efficient top-n recommendation for very large scale binary rated datasets (2013, background): “…Previous works (see [8] for example) pointed out that the conditional probability measure of similarity, P ( i | j ) , has the limitation that items which are purchased frequently tend to have higher values not because of their co-occurrence frequency but instead because of their popularity.”
Improving recommendation lists through topic diversification (2005, background): “Item-based CF [13, 26, 5] has been gaining momentum over the last five years by virtue of favorable computational complexity characteristics and the ability to decouple the model computation process from actual prediction making.”
Sequential Recommendation with User Memory Networks (2018, methodology): “…global average for better interpretability [16]: • Hit-ratio (HR): Hit-ratio gives the percentage of users that can receive at least one correct recommendation, which has been widely used in previous work [16, 30]: where I ( x ) is an indicator function whose value is 1 when x > 0, and 0 otherwise.”
MyMediaLite: a free recommender system library (2011, methodology): “In the development of the library, we employ best practices for (not only) free/open source projects, e.g. keeping the code in a public (and distributed) version control repository, having regular releases (roughly one per month) and Library Duine MultiLens LensKit Mahout GraphLab SUGGEST MyMe”
A Survey of Collaborative Filtering Techniques (2009, background): “Another similarity measure is conditional probability-based similarity [62, 63].”
Collaborative Filtering Recommender Systems (2007, background): “Methods researchers use to detect attacks are not even sensitive enough to detect harsh attacks [32].”
Collaborative filtering recommender systems (2013, background): “(2000b) and Karypis (2001) Prediction-based: Based on candidate set of C, the Eq.”
A Harmonic Extension Approach for Collaborative Ranking (2016, positive): “In contrast to previous work on top-N recommender systems [8, 5], we discourage the use of Precision@K in the context of collaborative ranking for recommender systems, which measures the proportion of actually rated items out of the top K items in the ranked list of all the items in the data set.”
Applying associative retrieval techniques to alleviate the sparsity problem in collaborative filtering (2004, methodology): “However with a different type of dataset in which the number of items is small and the number of users is large, the itembased approach should have better performance, as reported in the literature [Karypis 2001; Sarwar et al. 2001].”
Fast Greedy MAP Inference for Determinantal Point Process to Improve Recommendation Diversity (2017, methodology): “We adopt an item-based recommendation algorithm [24] on the training set to learn an item-item PSD similarity matrix S.”
Temporal recommendation on graphs via long- and short-term preference fusion (2010, methodology): “In the following, we investigate the impacts of the parameters β and η respectively, and then compare the Hit Ratios of the four temporal algorithms: MS-IPF, TPRW, TItemKNN, and TUserKNN.”