The Open UniversitySkip to content

Behavior based adaptive call predictor

Phithakkitnukoon, Santi; Dantu, Ram; Claxton, Rob and Eagle, Nathan (2011). Behavior based adaptive call predictor. ACM Transactions on Autonomous and Adaptive Systems, 6(3) Art 21.

Full text available as:
PDF (Proof) - Requires a PDF viewer such as GSview, Xpdf or Adobe Acrobat Reader
Download (1MB) | Preview
DOI (Digital Object Identifier) Link:
Google Scholar: Look up in Google Scholar


Predicting future calls can be the next advanced feature of the next-generation telecommunication networks as the service providers are looking to offer new services to their customers. Call prediction can be useful to many applications such as planning daily schedules, avoiding unwanted communications (e.g. voice spam), and resource planning in call centers. Predicting calls is a very challenging task. We believe that this is an emerging area of research in ambient intelligence where the electronic devices are sensitive and responsive to people’s needs and behavior. In particular, we believe that the results of this research will lead to higher productivity and quality of life. In this article, we present a Call Predictor (CP) that offers two new advanced features for the next-generation phones namely “Incoming Call Forecast” and “Intelligent Address Book.” For the Incoming Call Forecast, the CP makes the next-24-hour incoming call prediction based on recent caller’s behavior and reciprocity. For the Intelligent Address Book, the CP generates a list of most likely contacts/numbers to be dialed at any given time based on the user’s behavior and reciprocity. The CP consists of two major components: Probability Estimator (PE) and Trend Detector (TD). The PE computes the probability of receiving/initiating a call based on the caller/user’s calling behavior and reciprocity. We show that the recent trend of the caller/user’s calling pattern has higher correlation to the future pattern than the pattern derived from the entire historical data. The TD detects the recent trend of the caller/user’s calling pattern and computes the adequacy of historical data in terms of reversed time (time that runs towards the past) based on a trace distance. The recent behavior detection mechanism allows CP to adapt its computation in response to the new calling behaviors. Therefore, CP is adaptive to the recent behavior. For our analysis, we use the real-life call logs of 94 mobile phone users over nine months, which were collected by the Reality Mining Project group at MIT. The performance of the CP is validated for two months based on seven months of training data. The experimental results show that the CP performs reasonably well as an incoming call predictor (Incoming Call Forecast) with false positive rate of 8%, false negative rate of 1%, and error rate of 9%, and as an outgoing call predictor (Intelligent Address Book) with the accuracy of 70% when the list has five entries. The functionality of the CP can be useful in assisting its user in carrying out everyday life activities such as scheduling daily plans by using the Incoming Call Forecast, and saving time from searching for the phone number in a typically lengthy contact book by using the Intelligent Address Book. Furthermore, we describe other useful applications of CP besides its own aforementioned features including Call Firewall and Call Reminder.

Item Type: Journal Item
Copyright Holders: 2011 ACM
ISSN: 1556-4703
Project Funding Details:
Funded Project NameProject IDFunding Body
Not SetNot SetNSF
Extra Information: 28 pp.
Keywords: prediction; behaviour; call logs; call matrix; trace distance; convergence time
Academic Unit/School: Faculty of Science, Technology, Engineering and Mathematics (STEM) > Computing and Communications
Faculty of Science, Technology, Engineering and Mathematics (STEM)
Item ID: 35025
Depositing User: Santi Phithakkitnukoon
Date Deposited: 06 Nov 2012 09:23
Last Modified: 08 Dec 2018 10:20
Share this page:


Altmetrics from Altmetric

Citations from Dimensions

Download history for this item

These details should be considered as only a guide to the number of downloads performed manually. Algorithmic methods have been applied in an attempt to remove automated downloads from the displayed statistics but no guarantee can be made as to the accuracy of the figures.

Actions (login may be required)

Policies | Disclaimer

© The Open University   contact the OU