ISSS608 2017-18 T3 Assign Hasanli Orkhan Insights

Detecting Suspicious Individuals

Overview

Methodology

Findings

Conclusion

Changes over time in the company.Is the company growing?

• 

By looking at the overall trend we can compare absolute numbers and see that during 2.5 years the number of Emails always were higher than other communication types and average number of emails was 462K, whereas Calls which following the Emails on average showing 336K throughout the period. There is some obvious fluctuations depending on number of days in each month, we can see that for Emails, Calls and Purchase line plots there is a bit spike in months that have 31 days, a slight drop in months that have 30 days and reasonable drop in February of 2016 and 2017.
For meetings on the other hand there is an upward trend. Need to be noted that some months of meetings were filtered out because they showed values around 0. Starting from October 2015 we might notice increase in the number of meetings.

To compare the months in each year I have plotted the cycle plot where we might observe some drops and ups over the 2.5 years.

• 

By looking at this plot we may easily determine which months show drops or increase in the number of calls, emails or purchases relatively to other years. Again here for some months we only have 2016 and 2017 years and drop in February from 2016 to 2017 might be explained by 1 day difference.
Overall for months where we noticed increase it is obvious that increase from 2015 to 2016 relatively higher than increase from 2016 to 2017 which might indicates relatively slow increase in last year. Although for Emails and Calls the trend by months showing almost the same pattern, Purchases on the other hand differs by showing decrease from April to June, in October and December which is below average number for that months. It worth to mention that for December for all communications type there is a drop, however for Purchases it is obvious the percentage of the drop is quite high.
One of the suspicious patterns that we can extract by visualizing all the types of communications in one dashboard the highest months for Meetings happened in October and December 2017 when exactly purchase amount decreased substantially.

Who else appears to be closely associated with suspicious group?

After joining the suspicious data with large data set and running Modularity we have grouped individuals into 6 different communities. Individuals without 'Ins' adding were identified as new suspicious members.Below are shown all the members of the communities:

• 

  Community_1
• 

  Community_2
• 

  Community_3
• 

  Community_4
• 

  Community_5
• 

  Community_6

By Filtering only purchase type we can find out the names of the employees who is making suspicious purchases according to the insider information.

• 

From here we can see that from the suspicious group of employees Rosalia Larroque making purchase from Jenice Savaria.
There is only one person Gail Feindt who made the purchase with four suspicious individuals: Richard Fox, Tobi Gatlin, Meryl Pastuch and Lizbeth Jindra.

• 

Which employees are making suspicious purchases?

Organizational structure and a full picture of communications

Changes in the group composition depending on activities

Group interactions over time