Churn prediction in telecom industry using machine learning ensembles with class balancing
- Authors: Chowdhury, Abdullahi , Kaisar, Shahriar , Rashid, Md Mamunur , Shafin, Sakib , Kamruzzaman, Joarder
- Date: 2021
- Type: Text , Conference paper
- Relation: 2021 IEEE Asia-Pacific Conference on Computer Science and Data Engineering, CSDE 2021, Brisbane, 8-10 December 2021
- Full Text: false
- Reviewed:
- Description: Telecommunication service providers are going through a very competitive and challenging time to retain existing customers by offering new and attractive services (e.g., unlimited local and international calls, high-speed internet, new phones). It is therefore imperative to analyse and predict customer churn behaviour more accurately. One of the major challenges to analyse churn data and build better prediction model is the imbalance nature of the data. Customer behaviour for churn and non-churn scenarios may contain resembling features. Using a single classifier or simple oversampling method to handle data imbalance often struggles to identify the minority (churn) class data. To overcome the issue, we introduce a model that uses sophisticated oversampling technique in conjunction with ensemble methods, namely Random Forest, Gradient Boost, Extreme Gradient Boost, and AdaBoost. The hyperparameters of the baseline ensemble methods and the oversampling methods were tuned in several ways to investigate their impact on prediction performances. Using a widely used publicly available customer churn dataset, prediction performance of the proposed model was evaluated in term of various metrics, namely, accuracy, precision, recall, F-1 score, AUC under ROC curve. Our model outperformed the existing models and significantly reduced both false positive and false negative prediction. © IEEE 2022.
Dynamic content distribution for decentralized sharing in tourist spots using demand and supply
- Authors: Kamruzzaman, Joarder , Karmakar, Gour , Gondal, Iqbal , Kaisar, Shahriar
- Date: 2017
- Type: Text , Conference proceedings
- Relation: 13th IEEE International Wireless Communications and Mobile Computing Conference, IWCMC 2017; Valencia, Spain; 26th-30th June 2016 p. 2121-2126
- Full Text: false
- Reviewed:
- Description: Decentralized content sharing (DCS) is emerging as an important platform for sharing contents among smart mobile device users, where devices form an ad-hoc network and communicate opportunistically. Existing DCS approaches for tourist spot like scenarios achieve low delivery success rate and high latency as they do not focus on dynamic demand for contents which usually vary considerably with the number of visitors present or occurrence of some influencing events. The amount of available supply also changes because of the nodes leaving the area. Only way to improve content delivery service is to distribute the contents in strategic positions based on dynamic demand and supply. In this paper, we propose a dynamic content distribution (DCD) method considering dynamic demand and supply for contents in tourist spots. Simulation results validate the improvement of the proposed approach. © 2017 IEEE.
- Description: 2017 13th International Wireless Communications and Mobile Computing Conference, IWCMC 2017
Content exchange among mobile tourists using users' interest and place-centric activities
- Authors: Kaisar, Shahriar , Kamruzzaman, Joarder , Karmakar, Gour , Gondal, Iqbal
- Date: 2015
- Type: Text , Conference paper
- Relation: 2015 10th International Conference on Information, Communications and Signal Processing (Icics); Singapore, Singapore; 2nd-4th December 2015 p. 1-5
- Full Text: false
- Reviewed:
- Description: In this work we investigate decentralized content exchange among tourists who are mostly strangers, depicts irregular movement patterns and most likely not to have any prior social relationship or difficult to establish any in a tourist spot. We incorporate user's interest, trustworthy online recommendations, and place-centric information to facilitate content exchange in such tourist destinations. The proposed administrator selection policy considers stay probability in activities, connectivity among nodes and their available resources. We have done extensive simulation using network simulator NS3 on a popular tourist spot in Australia that provides a number of activities. Our proposed approach shows promising results in exchanging contents among users measured in terms of content hit and delivery success rate as well as latency. The success rate is comparable to those reported in the literature for cases where social relationship exist and nodes follow regular predictable movement patterns.
Content sharing among visitors with irregular movement patterns in visiting hotspots
- Authors: Kaisar, Shahriar , Kamruzzaman, Joarder , Karmakar, Gour , Gondal, Iqbal
- Date: 2015
- Type: Text , Conference paper
- Relation: 2015 IEEE 14th International Symposium on Network Computing and Applications (NCA); Cambridge, United States; 28th - 30th September 2015; published in Proceedings - 2015 IEEE 14th International Symposium on Network Computing and Applications, NCA 2015 p. 230-234
- Full Text: false
- Reviewed:
- Description: Smart mobile devices have become immensely popular among the people worldwide and provide a new platform for generating and sharing contents. The centralized and hybrid architectures for content sharing require constant Internet connection, increase traffic and incur costs. To address these issues several content sharing approaches have been proposed using the decentralized architecture. Most of the proposed approaches use spatio-temporal regularity and pre-existing social relationships of the users to predict their movements and facilitate content sharing. However, there are scenarios such as visiting hotspots where regular movement patterns or established social relationships among people might not exist. Content sharing in such scenarios has not been addressed yet in literature and existing prediction based approaches are ineffectual. This study focuses on facilitating content sharing in the afore-mentioned scenarios. We take account of user interests, recommendations from on-line social networks, hotspot specific activities and other relevant information to construct communities which facilitate content sharing. For each community an administrator, who maintains content and member lists and render directory services, is selected based on stay probability, interest score, battery lifetime and device configuration. Simulation results show that our proposed approach attains high content hit and success rate and low latency in delivery which is nearly comparable to those proposed for scenarios with regular predictable movement patterns reported in literature.