Numerical Data Where a system is capable of representing user inputs or actions as numerical data, it is useful to take into account whether these values
are accurate, whether they are commensurate, and whether they are independent.
Functions such as the geometric mean have a higher rate of change when input values are high than the arithmetic mean. This can help provide granularity
to the outputs, however it also means that errors on this portion of the domain
will influence the recommendation accuracy. In CF, two users might have similar preferences however one may consistently overrate items. In these cases, it
might make sense to standardize the ratings before aggregating so the values between users are comparable. The use of the WAM implies independence between
inputs, however other averaging functions, especially the Choquet integral, can
express interaction and correlation either among certain inputs or relative scores.
Numerical Data Where a system is capable of representing user inputs or actions as numerical data, it is useful to take into account whether these values
are accurate, whether they are commensurate, and whether they are independent.
Functions such as the geometric mean have a higher rate of change when input values are high than the arithmetic mean. This can help provide granularity
to the outputs, however it also means that errors on this portion of the domain
will influence the recommendation accuracy. In CF, two users might have similar preferences however one may consistently overrate items. In these cases, it
might make sense to standardize the ratings before aggregating so the values between users are comparable. The use of the WAM implies independence between
inputs, however other averaging functions, especially the Choquet integral, can
express interaction and correlation either among certain inputs or relative scores.
การแปล กรุณารอสักครู่..