تخصيص تحليل الشبكات الاجتماعية لشركات الاتصالات
Customize Social Network Analysis for Telecommunications Companies
اكثم فهيم عقل صوان
aktham faheem aqel sawan
MetadataShow full item record
Social Network Analysis (SNA) is created to analyze social network data. Therefore, the main companies in the data mining filed (such as IBM, SAS, R and python) have created their own SNA algorithms. The aim of this research is to create customized SNA algorithm for telecom companies because the current algorithms were not designed just for the telecom networks, in addition when current algorithms were used for telecom many high value customer not include in final result plus results coverage just 55% from input customers, so in the new algorithm relation strength and extenders were used to enhance final results 300 million records that belong to around 4 million customers in the last three were collected from (Jawwal Telecommunications Company) as case study. The current algorithms and the new algorithm were used the same data. In this research six experiments were applied based on call duration, call count and ratio between call duration and call count, in addition two groups size were used (15 and 20), Oracle Sql-PL/SQL was used to implement algorithm. The results that approved by Jawwal were based on parameters that used in experiment number six (ratio between calls count and call duration with group size till 20 customer), it has increased the coverage of NW to be 75.9% instead of around 55% for current algorithms, in addition all high valued customers has included in results for the new algorithm, moreover algorithm have applied in Mobily in Saudi Arabia and the same positive results have been found same as Jawwal. New novelty ideas have created in this research such as, extenders this type of customers used for customer who is influencer in one group and follower in the other group. Also relation strength used to create groups and assign followers to their most related influencer; furthermore, Super Group used as new layer to connect related groups in one group and find super influencer.
- Computer Science