How to recommend more items with high rating

First of all thank you for this great framework. I am working on a Restaurant/Food recommendation system with yelp dataset.
I have **item_features** like the category of the business, the number of reviews etc. I also included the average rating of each business. I have categorize rating : ["very_low, "low", "average","high"]. When i use the **dataset.build_item_features** function i use this form:
(item id, {feature: feature weight}) provided at documentation. As weight i use the actual value of rating (1,1.5,2,...5). For the other features i use weight: 1. So my question is how can i manage the weights so that businesses with low review value to have also lower score? One other idea is to pre-filter or post filter the items using in predict function and exclude those with rating under a specific threshold, but i would prefer to make a more "clever" recommender instead doing that.
About the **user_features** I am using the categories which user already have reviewed, with weight value to be (total reviews user did for that category)/(user's total reviews).
Apart from that i would also like to know if there is a way to give some more input in the predict function. For example if there are features such as Pizza, and Italian at item_features can i get predictions based on that or the only way is the post filtering?
Thanks in advance

Start with the documentation and API behavior for dataset.build_item_features and predict, focusing on how item and user feature weights are interpreted. Clarify whether rating thresholds, review-count weighting, and feature-based prediction inputs are supported requirements; done should be a concrete, documented scope or a narrowly defined feature request.

Written by the indexing model from the issue text.