Methods, apparatus, and prediction churn units are provided for remotely sensing and estimating and determining subscriber churn behaviour in a telecommunication network using a subscriber influence analysis. The analysis is based on a summarised weighted direct call graph of subscribers and call detail record (CDR) information between subscriber pairs iterated over a set of time periods. The influence of subscribers to each other is analysed by obtaining said CDRs belonging to subscribers of the telecommunication network to form the call graph. After forming said call graph, an influence map which determines how much each subscriber affects others is obtained using Call Duration (CD), Drop Count (DC), Call Count (CC) and SMS Count call indicators regarding the communication each subscriber makes with its neighbouring subscribers others over the telecommunication network.
展开▼
