Page 100 - Kỷ yếu hội thảo khoa học lần thứ 12 - Công nghệ thông tin và Ứng dụng trong các lĩnh vực (CITA 2023)
P. 100

84


                       An advantage of median is that it is not sensitive to outliers. This can be especially
                     relevant in cases where respondents answer without any relevant knowledge or if they
                     do not give much thought to their judgment [8].


                     4.4   Weighted aggregation

                     Weighted  aggregation  is  a  technique  used  in  prediction  that  involves  assigning  a
                     weight or importance factor to each individual's prediction. The weight is determined
                     based on several factors such as their credibility, expertise, or past performance. In
                     the study described in [21], the authors assigned weights based on self-assessment,
                     knowledge, and hit rate in their research. On the other hand, Nguyen VD et al. in [22]
                     assigned  weights  based  on  the  individual's  reputation.  Some  studies  used  different
                     methods  such  as  the  accuracy  of  recent  judgments  [23],  or  summing  probability
                     predictions [24].
                       Weighting the crowd produced both positive and negative outcomes. When relying
                     on the confidence of individual members, the crowd did not exhibit wisdom, as their
                     personal biases led to overconfidence [12]. However, von der Gracht et al. [21] found
                     that the individuals were not overconfident and exhibited more wisdom compared to
                     individuals  aggregated  by  their  past  performance.  Despite  the  favorable  outcomes,
                     they  did  not  discover  any  added  benefits  in  weighting  the  crowd  since  equally
                     weighting  the  responses  resulted  in  superior  outcomes.  In  their  study  presented  in
                     [25],  the  authors  demonstrated  that by removing underperforming  individuals  from
                     the  group  and  only  considering  those  who  made  a  positive  contribution  (model
                     CWM), the overall performance of the crowd (which is highlighted in bold in Fig 3)
                     improved  significantly. This approach  was  found to be more  effective than relying
                     solely on past performance for weighting.















                             Fig. 3. Performance of the Models Compared (in Terms of Their Scores) [25]


                     Overall,  the method of assigning weights to individuals depends on various factors
                     such as the type of data being analyzed and the objectives of the research.


                     4.5   Mode

                     The mode can be a useful measure of central tendency when dealing with categorical
                     data, such as  words  or  other non-numerical values. For example, it  can be used to
                     determine which football player or team performs best [26], or which political party
                     will  win  the  election  [27].  In  cases  where  the  data  is  categorical,  it  may  not  be




                     CITA 2023                                                   ISBN: 978-604-80-8083-9
   95   96   97   98   99   100   101   102   103   104   105