Search code examples
datasetwekarecommendation-engine

Dataset size for recommendation algorithm


I am working on creating a recommender system,using weka. What i read about dataset on internet so far is It should have more data for recommendation to be good. I wonder what if I am trying to recommend something about which we can't have large amount of data. For example if want to create a Park or hospital recommender system for a particular city. It wouldn't have more than 100 entries. So is it ok to do it with small datasets?


Solution

  • As reported in this post, there are some guides or rules-of-thumb that could be used for preparing sufficient data for the model. Some of the advice here included:

    • At least six times the number of features as training cases
    • At least three times the number of features per class

    Please note that these are only rules of thumb, and may or may not be enough depending on the complexity of your data. Based on the recommendations outlined in this article, 100 cases might just be enough, but as Ramanan R stated, more cases would likely be better.