One of the metrics to evaluate Recommender systems is Predictive accuracy metrics, I want to know what is the accepted value for it is 60 % or 70 % is accepted?
Nouran Radwan
Train And Test Validation
- Cross validation
Accuracy Metrics
- Root Mean Square Error
- Mean Absolute Error
https://medium.com/the-owl/evaluating-recommender-systems-749570354976
https://towardsdatascience.com/evaluation-metrics-for-recommender-systems-df56c6611093
Hope this is helpful.
Dear Nouran Radwan ,
Predictive accuracy metrics, classification accuracy metrics, rank accuracy metrics, and non-accuracy measurements are the four major types of evaluation metrics for recommender systems.
https://analyticsindiamag.com
Regards,
Shafagat
I didn't want to know the types of evaluation metrics but I want to know the accepted percentage.
Nouran Radwan I believe your question is not specific enough to be answered. There is nothing called accepted accuracy as it depends on similarity measure used with KNN, or ML model. If you want to take a deep knowledge into similarity measures based RS, I advise you to go through my publications:
Article Enhancing recommendation systems performance using highly-ef...
Preprint Combinations of Jaccard with Numerical Measures for Collabor...
Dear Nouran Radwan ,
As I understood that you don't want the metrics for evaluating the recommender system (RS) but you want the specific value for evaluating the RS (if the value of the metric is enough to say the RS is good or no).
To begin, you should be aware that there is no specific percentage for evaluating RSs. The percentage of the RS's metrics will vary depending on your domain of work, similarity metrics, prediction algorithms, and recommendation methods. Each time u use different algorithm or method or similarity metric or u change the domain, the value of the metric will also change.
The correct way to evaluate your RS is to compare your metrics' values with another work that employs similar ideas or methods as yours. Then, check the difference between you metrics' value and the other work values. The difference will decide the efficacy of your work.
Good Luck!
Check https://analyticsindiamag.com/how-to-measure-the-success-of-a-recommendation-system/
https://towardsdatascience.com/evaluation-metrics-for-recommender-systems-df56c6611093
If something is more than 50% is acceptable, so to me I think the two values are okay and I shall go for 70% because it is higher.
Dear Nouran Radwan,
You can find the paper related to RS metrics -> Article Experimental interpretation of adequate weight-metric combin...
Best Regards
If we talk about accuracy the higher is the better But acceptance depends on its use case.
- Badges
- Science topic