56th event Ranking General algorithm 4 steps

Since following the master of the study, also fell in love with reading, more curious, the major e-commerce website book ranking in the end is based on what factors or dimensions to arrange? What sort of ranking algorithm is more scientific? Is there a more general algorithm? Master must have a way. Really, consulting Master, learned that there is a general ranking algorithm.

The general-purpose leaderboard algorithm is divided into four steps: The first step is to clearly influence the dimensions or factors of the leaderboard; the second step standardizes the numerical values of each dimension or factor; The third step is to determine the weight ratio of each dimension or factor; The fourth step calculates a comprehensive score after ranking.

Let us elaborate on a concrete example. Suppose there are now 20 apples for the user to eat or try, the rules of the game is the head of friends or friends in the circle together to rob Apple's eligibility. Assuming that there are now three heads involved in the group, we follow the above four-step one by one disassembly:
The first step: whether the group can obtain a disqualification, the main dimensions or factors are: groups have not been eligible for the number of users, the group before the regiment to obtain the number of users who have been eligible and reviewed, the total number of groups, the team submitted time, the leader level (the virtual points obtained).
The second step: the numerical standardization of each dimension or factor, standardization mainly has max-min standardization and z standardization, we use the former, the formula is (x-min)/(Max-min), the group submission time adopts (max-x)/(Max-min), Because the sooner the group submission time, the easier it is to qualify for a disqualification. Standardized scoring = All data standardized scoring *100, which translates into percentile.
Step three: Determine the weight ratio of each dimension or factor. So how do you know? Assuming that there are ABCD four dimensions, we determine the weight of the steps: The 1th step is simplified to 1/0-type logical thinking, the 2nd step expert votes, the longitudinal axis is more important than the axis fill "1", instead of "0"; the 3rd step adds each line of numbers, sorted by the sum of the values, as shown in matrix Analysis 7-18.

Because there may be a value of 0 in the total (compared to row d of the table shown in 7-18), for ease of calculation, we think that the sum of each of the values is +1.

The weight of weight of =3/10x100%=30%,b is obtained by the formula of the weight of =4/10x100%=40%,c =2/10x100%=20%,d weight =1/10x100%=10%.
Of course, according to the actual needs of the weight than can be customized configuration, so more flexible.
Fourth Step: Comprehensive score = The standardized score of the number of non-people in the group * corresponding weight + standardized score of group review number * corresponding weight + standardized score for total number of groups * corresponding weight + standardized score for group submission time * Corresponding weights + leader level standardized score * corresponding weights, The head of the leaderboard is shown in final result 7-19.

Just after listening to the tips of the master, vaguely feel the use of some previously learned data analysis knowledge and skills. Epiphany, the original data standardized scoring and the method of determining weight matrix analysis were used synthetically.

Product managers do not learn things to learn things, the real understanding, a lot of complex problems can be solved, it is the use of data standardization scoring and determine the weight of the matrix analysis method, a synthesis of the leaderboard is a general algorithm. When it comes to ranking, it's easy to get a ranking result by simply changing the dimensions, which is what you really do.

56th event Ranking General algorithm 4 steps