1
E-commerce Website Analysis based on MCDM Methods
Madison Bright, Michelle McGowan, Sakethva Pasumarti
2
Our Team
3
INTRODUCTION
4
4/30/2018
MCDM Explained
Identify Criteria
Assign Weightage
Generate Alternatives
Evaluate & Analyze
5
Process
Literature Review
AHP + TOPSIS + Maybac
Data Collection
Analysis & Code
6
Websites and Why
Amazon - most popular globally
Ebay - one of the first
Aliexpress - Asia’s leading site
Temu - emerging site
7
Criteria
Loading Time - too slow can cause customers to leave
Security Risks - giving financial information
Usability - easy to use
Page Rank - top 10
8
Criteria
Keywords - right words show up at the right time
Traffic - more people more trustworthy
Server Response Time - access to the website
Key Literature Overview
9
10
11
Tools
Traffic & Page Rank - Similarweb
Load Time & Usability - Solarwinds pingdom
Security - Securi
Keywords - ahrefs
Server Response Time - ManageEngine Site24x7
12
4/30/2018
Alternative Tools
13
Data Collection
Collect
Check
Average
The method to collect the data was straightforward. We collected data daily at random times so we could collect a wide range of data. Based on the tools we used, we were given a metric as either a number, score , or another type of value.
All of the tools we used were web based therefore we manually collected the data everyday. This allowed us to monitor the tools to ensure that data was taken from the same servers.
Although the data was vast in nature, we later condensed it into an averaged version for our calculations.
14
Data Collection Graphs
The data was collected using the aforementioned tools in a quantitative and qualitative manner. Some values were changing daily while others remained fairly constant.
This is a small portion of the raw data collected over the span of two weeks. It was collected at different times to mimic randomization.
15
The above table is the master table. This includes the units for each criteria, whether a minimum or maximum value is more beneficial, and the averages for each column.
Data Collection Graphs Cont.
The table to the right depicts the an example of our qualitative data. In the master table the qualitative values have been converted to values on a scale from 1-9.
1
2
16
Methods for Evaluations
AHP
TOPSIS
MABAC
This method allows us to weight the criteria for the other two methods.
This allows us to use vector normalization to normalize our qualitative and quantitative data.
This method uses border approximation area to limit decision maker inconsistency.
17
AHP
AHP
To help in decision-making, the Analytical Hierarchical Process (AHP), directs users to make deliberate and logical decisions. Thomas Satty created the method in 1981. It was created to assist decision-makers in selecting the optimal choice based on multiple criteria. When it comes to several criteria and alternatives, AHP is quite beneficial, and alternatives can benefit from a pairwise comparison. This method determines the weight for each criterion, chooses the top pairs, then compares them on a scale of 1 to 9.It is determined which of the two entities has a bigger amount by comparing the two. The decision is made after the criteria have been calculated.
Method 1
18
This table is the same as the master table without the extra information about units and min/max.
AHP Weight Calculation
This matrix is where the decision maker can input the weights for each criteria. We compare the rows versus the columns in the upper triangular matrix. Then we assign a score from 1-9 or 1/9 - 1 based on the relative importance. Finally the values are inverted to the lower triangular matrix and the column totals are calculated.
19
AHP Weight Calculation Cont.
Each cell is divided by the total for each column. This gives us a weighted value for each cell. Then, the weights for the criteria are determined by averaging each row. Finally the criteria can be ranked to determine which is the most important.
20
TOPSIS
TOPSIS
TOPSIS is a decision making tool that was developed in 1981 by Hwang and Yoon. This method relies on vector normalization and ideal separation. AHP generates the weights for each of the criteria and TOPSIS uses these as a part of the normalization process. From there, the weighted normalized values, the VJ+, and the VJ- values can be solved for. Finally, the ideal separation can be used to calculate the confidence levels. The closer the confidence level is to 1, the higher the ranking will be.
Method 2
21
TOPSIS Calculation
Here the vector normalization is done. First the original values from the master table are squared, then summed by column. Then each value from the original columns is divided by the sqrt(sum) value for the respective column. This gives us the vector normalized values for each cell.
Formula for Vector
Normalization
22
TOPSIS Calculation
The Weighted normalized values are found by multiplying each cell in the Vector Norm column by its category weight. Then the VJ+ and the VJ- values indicate the best and the worst value from each column. The categories are labeled B and NB (beneficial and non-beneficial). For the beneficial ones it is better to have a larger value and vice versa for non-beneficial.
23
TOPSIS Calculation
Finally, the Ideal Separation or Euclidean distance can be calculated based on the formula shown. The values for DI+ are found by subtracting the weighted norm values from the VJ+ value. The values for DI- are found by subtracting the weighted norm values from VJ-. Finally, the confidence level can be found by dividing the DI- values by the sum of the DI+ and DI- values in each row. The highest CL value receives the highest ranking.
Formula for Euclidean Distance
24
MABAC
MABAC
MABAC was originally created by Pamucar and Cirovic in 2015. This method is fairly new however it has an interesting nuance compared to TOPSIS. MABAC uses border approximation area which essentially allows the user to limit decision maker error. This error can range from inconsistencies with ranking during the weightage process to general data errors.
Method 3
25
MABAC Calculation
Here the same master table as the TOPSIS method is used. The beneficial and non-beneficial criteria are also labeled. The x+ value indicates the best value in the column and the opposite for the x- value. The max-min is the x+ value minus the x- value and vice versa for the min-max value. These will be used in the next step.
26
MABAC Calculation
The normalization table is calculated differently based whether the category is beneficial or non-beneficial. For beneficial values we subtract the cell value from the x- value and then divide by the max-min value. For non-beneficial values we subtract the cell value from the x+ value and then divide by the min-max value. Finally the weighted norm is found by multiplying the normalized value by the category weight, then adding the category weight.
Normalization
Formulas
27
MABAC Calculation
Here the BAA (Border Approximation Area) is calculated by taking the product of each column and raising that value to the power of 0.25. The V-G matrix calculates the distance of the values. The maximum value is the best in the category and the smallest value is the worst. A negative value entails that the website performed subpar in that category. Finally the values are summed across the rows. The highest value receives the highest ranking.
BAA
Formula
28
Evaluation Results
A
E
AL
T
29
Evaluation Results
30
Solutions
31
Lorem ipsum dolor sit amet, consectetur adipiscing elit. Phasellus non nisi accumsan, pretium ligula quis, iaculis erat. Nulla elit erat, pulvinar a erat ac, tincidunt consequat leo.
Lorem ipsum dolor sit amet, consectetur adipiscing elit. Phasellus non nisi accumsan, pretium ligula quis, iaculis erat. Nulla elit erat, pulvinar a erat ac, tincidunt consequat leo.
What Next?
32
Acknowledgments
Dr. Dhanapal Durai Dominic
Dr. Bimal Nepal
Dr. Om Prakash Yadav
Dr. Eakalak Khan
Ms. Iffah
Mr. Khairul �Student Mobility Office
This work was supported by the National Science Foundation’s International Experience for Students (IRES) Site grant. (Grant Numbers: OISE# 1952490-TAMU, 1952493-NC A&T State, and 195249-UNLV7). Any opinions, findings, conclusions, or recommendations presented are those of the authors and do not necessarily reflect the views of the National Science Foundation. We also appreciate the work of Dr. Jessica Martone’s team from The Mark USA in providing the evaluation data for this project.
33
THANK YOU