Trying to find peer-reviewed papers about what method of averaging is the most ideal (Harmonic, Geometric, Arithmetic, Quadratic)

I am currently working with a Netflix Dataset for research, and I needed to find papers in regards to which method of averaging is most ideal for this situation. For example, We deal with ratings of movies, and we want to guess how people would rate movies based upon the ratings of their friends or neighbors. If we cannot find a rating that is definitive, we need to set an 'average' or default for that rating until we can do so. My colleagues and I were thinking of using the Harmonic mean, since the variance of ratings for each individual could be great, but at the same time an individual could have ratings that change very little. Are combinations of averages better than, as opposed to just one? I tried to be as descriptive as possible, any form of published paper that could help me along would be greatly appreciated!