Socioeconomic mapping using mobile call detail records for Sri Lanka

Abstract

CDR (Call Detail Record) is a data record that is generated by a telephone exchange or telecommunication equipment which contains details of that telephone call. These records are utilized by telecommunication service providers for their billing purposes. High volume of data generates in quick time which contains customer specific data with temporal and geographic information. Other than CDR data, telco systems have various data sources such as customer payment data and device information. Telco service providers collect CDR and store them for a limited period of time for various activities. It can be repurposed other than billing activities. CDR data can denote various aspects of human behavior such as human relationships, expenditure power and mobility. Those aspects can help governance of the country regarding economic development and resource allocation in timely manner. In this research, CDR data records were integrated with other telco data sources in order to analyze and predict the economic behavior of a specific geographical area in Sri Lanka. Big data and Machine Learning techniques were used to extract the customer behavior from CDR data. Big data processing techniques were applied on CDR data and telco data sources in order to identify properties of customers in a specific geographic area over a time period. Then those identified properties were evaluated to see whether they reflect the economic behavior in that area or not. After identifying dominant features related to the economy, Machine Learning techniques were applied on them to see the feasibility of predicting the economic behavior in the targeted area. The results were evaluated and interpreted as a part of this research. Such results will be very useful for the governance in order to understand the economic conditions in a specific geographical area and make the policies to address poverty over the time.