One could argue that this behavior is actually a good thing -- basing an estimate of similarity based on one data point could be very unreliable.
There are some similarity metrics that don't have this property, but they all basically ignore the rating value. See LogLikelihoodSimilarity or TanimotoCoefficientSimilarity for instance. You could try them, but, if you have ratings I would think you want to start by using them. In that case, I'd suggest you make up some more data instead! On Tue, Jun 23, 2009 at 6:43 PM, charlysf<[email protected]> wrote: > > Of course, thank you very much, it's because now my table is almost empty, in > this case, do you recommend to me to use an other similarity ?
