CODEWITHSUNDEEP
mathstatisticsclassificationstatistical-test
elizabeth
elizabeth

Reputation: 73

Aggregation Method in Classification

There is a set of objects to be classified into n classes. Each class has its own coefficient.

For example:

class      coefficient
 1             0.5
 2             0.7
 N             0.4

I arranged objects into classes and get the set of coefficients. But i need to determine a single coefficient that would be most suitable for the entire set of objects.

How can I Aggregate the Objects' Coefficients and choose one common coefficient? Except for the methods of average coefficients.

Upvotes: 1

Views: 37

Answers (2)

quester
quester

Reputation: 564

in python

  • pandas.value_counts over coefficients, and pick most common
  • collections.Counter over coefficients, and pick most common

or plain old histogram

other than that you can make something like weighted average

there are many statistics, depends on what you need

Upvotes: 0

user24714692
user24714692

Reputation: 4949

This depends on the nature of your data and any specific requirements of your problem.

You can perform variety of statistical analysis on the data and see which ones provide the best classification.

For instance, the median and mode are particularly useful when dealing with outliers. Harmonic and geometric means are suitable for multiplicative data, and there are numerous other methods to apply.


import numpy as np
from scipy import stats

def get_aggr(C):
    _med = np.median(C)
    _mod = stats.mode(C).mode
    _hmean = stats.hmean(C)
    _gmean = stats.gmean(C)
    _trim_mean = stats.trim_mean(C, 0.1)
    return _med, _mod, _hmean, _gmean, _trim_mean


C = [0.5, 0.7, 0.4, 0.5, 0.6, 0.4]
print(get_aggr(C))

Prints

(0.5, 0.4, 0.49606299212598426, 0.5060788802128566, 0.5166666666666666)

Upvotes: 0

Related Questions