Fierce competition in many industries causes switching behavior of customers. Because foregone profits of defected customers are significant, an increase of the retention rate can be very profitable. In this paper, we...
详细信息
ISBN:
(纸本)1853129259
Fierce competition in many industries causes switching behavior of customers. Because foregone profits of defected customers are significant, an increase of the retention rate can be very profitable. In this paper, we focus on the treatment of companies' most promising current customers in a non-contractual setting. We build a model in order to predict chum behavior of top clients who will (partially) defect in the near future. We applied the following classification techniques: logistic regression, linear discriminant analysis, quadratic discriminant analysis, C4.5, neural networks and Naive Bayes. Their performance is quantified by the classification accuracy and the area under the receiver operating characteristic curve (AUROC). The experiments were carried out on a real life data set obtained by a Belgian retailer. The article contributes in many ways. The results show that past customer behavior has predictive power to indicate future partial defection. This finding is from a companies' point of view even more important than being able to define total defectors, which was until now the traditional goal in attrition research. It was found that neural networks performed better than the other classification techniques in terms of both classification accuracy and AUROC. Although the performance benefits are sometimes small in absolute terms, they are statistically significant and relevant from a marketing perspective. Finally it was found that the number of past shop visits and the time between past shop incidences are amongst the most predictive inputs for the problem at hand.
In the marketing domain, sequential patterns have been usefully deployed for predicting various aspects of customer purchase behavior. However, to date, the applications of the technique have mainly focused on improvi...
详细信息
ISBN:
(纸本)1853129259
In the marketing domain, sequential patterns have been usefully deployed for predicting various aspects of customer purchase behavior. However, to date, the applications of the technique have mainly focused on improving algorithms for detecting sequentially related events, whereas the implications of the sequences, and their, incorporation into a global structure of consecutive sequences have been treated to a lesser extent. In this paper, such a structure, that we will refer to as sequential architecture, will be empirically investigated for a specific case in a fast moving consumer goods setting.. Hence, the goal of this paper was to introduce a new concept that might prove to be a;relevant tool for marketing decision making rather than offering a sound solution within a clearly demarcated problem definition. As opposed to the traditional sequence-analysis approaches, in this study, an array of binary logit analyses was applied for detecting significant sequences among category purchases. We use the output of the logit analyses to define the category that is most significantly influenced per newly purchased category, and we select these links for constructing the applicable sequential architecture. Finally, we provide empirical evidence that the methodology suggested is able to double the performance of predicting purchases in categories that were not purchased previously by the consumer, compared to a random model. In summary, we have shown that (i) binary logit analysis provides a feasible alternative for detecting and selecting highly significant sequential relationships, (ii) a sequential architecture can be successfully compiled through the methodology offered in this paper, and (iii) the provided sequential architecture can be a useful tool in understanding and predicting customer behavior. Future applications possibly lie ahead in the field of inter-category management, shelf-space allocation, store-layout decisions, retailer promotions, customer profiling
This paper describes work towards automatically building on-line structured information resources from information sources that are comprised largely of natural language but with some structuring conventions. Such con...
详细信息
When datamining first appeared, several disciplines related to data analysis, like statistics or artificial intelligence were combined toward a new topic: extracting significant patterns from data. The original data ...
详细信息
Determining the convex hull of a point set is a basic operation for many applications of patternrecognition, image processing, statistics, and datamining. Although the corresponding point sets are often large, the c...
详细信息
Research on recognition and generation of signed languages and the gestural component of spoken languages has been held back by the unavailability of large-scale linguistically annotated corpora of the kind that led t...
详细信息
Research on recognition and generation of signed languages and the gestural component of spoken languages has been held back by the unavailability of large-scale linguistically annotated corpora of the kind that led to significant advances in the area of spoken language. A major obstacle has been the lack of computational tools to assist in efficient analysis and transcription of visual language data. Here we describe SignStream, a computer program that we have designed to facilitate transcription and linguistic analysis of visual language. machine vision methods to assist linguists in detailed annotation of gestures of the head, face, hands, and body are being developed. We have been using SignStream to analyze data from native signers of American Sign Language (ASL) collected in our new video collection facility equipped with multiple synchronized digital video cameras. The video data and associated linguistic annotations are being made publicly available in multiple formats.
The proceedings contain 42 papers. The special focus in this conference is on Probabilistic Models and Estimation. The topics include: A double-loop algorithm to minimize the bethe free energy;a variational approach t...
ISBN:
(纸本)3540425233
The proceedings contain 42 papers. The special focus in this conference is on Probabilistic Models and Estimation. The topics include: A double-loop algorithm to minimize the bethe free energy;a variational approach to maximum a posteriori estimation for image denoising;maximum likelihood estimation of the template of a rigid moving object;an application to shape retrieval;a fast MAP algorithm for 3D ultrasound;designing the minimal structure of hidden markov model by bisimulation;relaxing symmetric multiple windows stereo using markov random fields;camera calibration for 3-D surface reconstruction;a hierarchical markov random field model for figure-ground segregation;articulated object tracking via a genetic algorithm;learning matrix space image representations;supervised texture segmentation by maximising conditional likelihood;optimization of paintbrush rendering of images by dynamic MCMC methods;illumination invariant recognition of color texture using correlation and covariance functions;path based pairwise data clustering with application to texture segmentation;a maximum likelihood framework for grouping and segmentation;image labeling and grouping by minimizing linear functionals over cones;grouping with directed relationships;segmentations of spatio-temporal images by spatio-temporal markov random field model;highlight and shading invariant color image segmentation using simulated annealing;edge based probabilistic relaxation for sub-pixel contour extraction;two variational models for multispectral image classification;an experimental comparison of min-cut/max-flow algorithms for energy minimization in vision;a discrete/continuous minimization method in interferometric image processing and a transformation approach.
We present a framework for visualizing remote distributed data sources using a multi-user immersive virtual reality environment. DIVE-ON is a system prototype that consolidates distributed data sources into a multidim...
详细信息
datamining is the science of finding unexpected, valuable, or interesting structures in large data sets. It is an interdisciplinary activity, taking ideas and methods from statistics, machinelearning, database techn...
详细信息
datamining is the science of finding unexpected, valuable, or interesting structures in large data sets. It is an interdisciplinary activity, taking ideas and methods from statistics, machinelearning, database technology, and other areas. It poses novel challenges, in part arising from the sheer size of modem data sets. Although there is no doubt that it addresses important questions, there are deep issues to be resolved relating to data quality and the nature of inference. Statisticians have an important role to play in resolving these issues.
暂无评论