Enhancing deep learning models for campaign participation prediction

Abstract

Companies engage with their customers in order to establish a long-term relationship. Targeting the right audience with the right product is crucial for providing better services to customers, increasing their loyalty to the company, and gaining high pro t. Therefore, companies make huge investments to build campaign management systems, which are mostly rule-based and highly depend on business insight and human expertise. In the last decade, recommendation systems usually use modeling techniques such as deep learning to understand and predict the interests of customers. Classic deep neural networks are good at learning hidden relations within data (generalization); however, they have limited capability for memorization. Wide & Deep network model, which is originally proposed for Google Play App. recommendation, deals with this problem by combining Wide & Deep network models in a joint network. However, this model requires domain expert knowledge and manually crafted features to bene t from memorization. In this thesis, we advocate using Wide & Deep network models for campaign participation prediction, particularly in the area of telecommunication. To deal with the aforementioned issue with that model, this thesis introduces the idea of using decision trees for automatic creation of combinatorial features (cross-product transformations of existing features) instead of demanding them from human experts. A set of comprehensive experiments on campaign participation data from a leading GSM provider has been conducted. The results have shown that automatically crafted features make a signi cant increase in the accuracy and outperform Deep and Wide & Deep models with manually crafted features. Furthermore, since a limited number of access to the customers is allowed, making well-targeted o ers that are likely to be acceptable by the customers plays a crucial role. Therefore, an e ective campaign participation prediction require to avoid falsepositive predictions. Accordingly, we extended our research towards classi cation uncertainty to build network models that can predict whether or not they will fail. Consequently, we adopt evidential deep learning models to capture the uncertainty in prediction. Our experimental evaluation regarding prediction uncertainty has shown that the proposed approach is more con dent for correct predictions while it is more uncertain for inaccurate predictions.Firmalar mu sterileri ile uzun vadeli ili skiler geli stirebilmek ad na ileti sim kurarlar. Do gru kitleyi do gru urunle hede emek, mu sterilere daha iyi hizmet sunarak sirkete olan ba gl l klar n art rma ve dolay s yla sirket gelirlerini art rma noktas nda kritik onem ta s r. Bu nedenle sirketler, co gunlukla kural bazl cal san ve insan uzmanl g na dayal kampanya yonetim sistemleri olu sturmak i cin buyuk yat r mlar yapmaktad r. Son on y lda oneri sistemleri genellikle mu sterilerin beklentilerini anlamak ve tahminlemek amac yla derin o grenme gibi modelleme tekniklerini de kullanmaktad r. Klasik derin sinir a glar veri i cindeki gizli ili skileri o grenmede olduk ca ba sar l d r (genelleme) ancak hat rlama noktas nda daha s n rl yetkinlikleri vard r. Aslen Google Play uygulamas i cin tasarlanan Wide & Deep o grenme modeli Wide ve Deep a g modellerini tek bir model i cerisinde birle stirerek bu soruna cozum sunmaktad r. Bununla birlikte, bu model, uzman bilgisine ve uzmanlar taraf ndan birden fazla verinin bir araya getirilmesi ile haz rlanm s girdilere ihtiya c duyar. Bu tezde, ozellikle telekomunikasyon alan nda, kampanya kat l m tahmini i cin Wide & Deep a g modellerini kullanmay oneriyoruz. Tez kapsam nda, modellerde uzman bilgisine duyulan ihtiyac ortadan kald rmak amac yla, ba g ms z de gi skenler aras ndaki capraz ili skileri otomatik olarak olu sturmak i cin karar a gaclar n kullanan bir yontem geli stirilmi s ve GSM mu sterilerinin kampanya kat l m n i ceren ger cek veri setleri uzerinde kapsaml bir de gerlendirme yap lm st r. Yap lan kar s la st rma sonucunda onerilen yontem ile Wide & Deep model performans n n onemli oranda art r labildi gi ve Deep, Wide & Deep modellerden daha iyi sonu clar elde edildi gi gozlemlenmi stir. Ayr ca, kampanya sistemlerinde sadece s n rl say da mu steri eri sime izin verildi ginden, kabul edilme ihtimali en yuksek olan tekli erin mu steriye sunulmas kritik onem ta s maktad r. Dolay s yla, kampanya kat l m n n ba sar yla tahminlenebilmesi hatal tahminlerden ka c nmay da gerektirmektedir. Bu nedenle ba sar s z olup olmayacaklar n tahmin edebilecek a g modelleri olu sturmak amac yla ara st rmam z s n and rma belirsizli gi konusunda geni sleterek kan ta dayal derin o grenme modellerini cal smam za dahil ettik. Sonu clar, onerilen modelin do gru s n and r lm s o geler uzerinde en du suk belirsizlik de gerine sahip oldu gunu, yanl s s n and rd g o geler uzerinde ise daha yuksek belirsiklikle karar vermi s oldu gunu gostermektedir.