Mobile phone Data

Features

Individual

• active_days
• number_of_contacts
• number_of_interactions
• call_duration
• percent_nocturnal
• percent_initiated_conversations
• response_delay_text
• response_rate_text
• entropy_of_contacts
• Interactions_per_contact
• percent_pareto_interactions (percentage of user’s contact that account for 80% of its interactions)
• percent_pareto_durations

Spatial

• Number of unique places (antennas) visited
• Entropy of antennas
• percent at home
• radius of gyration (the equivalent distance of the mass from the center of gravity, for all visited places)
• frequent_antennas - location that accounts for 80% of the locations the user was
• churn_rate - Computes the frequency spent at every towers each week, and returns the distribution of the cosine similarity between two consecutives week

Network (These are graph network analysis)

• Directed, weighted matrix for call, text etc
• Directed, Unweighted matrix
• Undirected, weighted matrix
• Undirected, Unweighted matrix
• Clustering coefficient - Measure of the degree to which nodes in a graph tend to cluster together
• clustering coefficient unweighted of users weighted undirected network
• clustering coefficient weighted (undirected)
• assortativity of indicators(The extent to which nodes of a graphlink to others of the same degree)
• assortativity of attributes

Recharges

• Recharge amounts
• Time between recharges
• percent pareto recharges
• Number of recharges
• Average daily balance estimated from all recharges

Challenges

• Mobile data will not be uniform across different networks. A different model may be required for different network.