Image analysis considering textual correlations enables accurate user switching tendency prediction

Abstract

Predicting likely-to-churn users employing surveys is a challenging task. Individuals with different personalities may make different choices in the same situation, so we introduced social media avatars that reflect the user’s psychological state when analyzing their churn tendency. In this paper, we propose a multimodal framework that jointly learns image and text features to establish correlations among users with low net promoter score (NPS) and those likely to churn. We conducted experiments on actual data, and the results show that our proposed method can identify NPS-degraded users in advance, promoting the commercial development of the operator.

References

[1] CooperA B, BlakeA B, PaulettiR E, et al.. Personality assessment through the situational and behavioral features of instagram photos[J]. European journal of psychological assessment, 2020, 36: 959-972

[2] RizomyliotisI, PoulisA, ApostolosG, et al.. Applying FCM to predict the behaviour of loyal customers in the mobile telecommunications industry[J]. Journal of strategic marketing, 2020, 28(1):1-15

[3] OUYANG Y, WANG L, YANG A, et al. The next decade of telecommunications artificial intelligence[EB/OL]. (2021-01-19) [2022-12-13]. https://arxiv.org/ftp/arxiv/papers/2101/2101.09163.pdf.

[4] HapsariR, HusseinA S, HandritoR P. Being fair to customers: a strategy in enhancing customer engagement and loyalty in the Indonesia mobile telecommunication industry[J]. Services marketing quarterly, 2020, 41(1):49-67

[5] HuangY Y, LiuY T, YuL M. Research on NPS survey data of telecom companies[J]. Post and telecommunications design technology, 2018, 509(07):52-56

[6] LvJ. Research on network optimization guided by NPs based on user network net recommendation value[J]. Data communication, 2019, 5: 8-17

[7] SarohaR, DiwanS P. Development of an empirical framework of customer loyalty in the mobile telecommunications sector[J]. Journal of strategic marketing, 2020, 28(8):659-680

[8] JeremyN H, ChristianG, KamalM F, et al.. Automatic personality prediction using deep learning based on social media profile picture and posts[C], 2022, Yogyakarta, Indonesia. New York, IEEE: 21563700

[9] ZhuH, ZhouY, LiQ, et al.. Personality modeling from image aesthetic attribute-aware graph representation learning[J]. Journal of visual communication and image representation, 2022, 89: 103675

[10] SomayeA, MohsenE M. Image recommender system based on compact convolutional transformer image style recognition[J]. Journal of electronic imaging, 2022, 31: 043054

[11] GATYS L A, ECKER A S, BETHGE M. A neural algorithm of artistic style[EB/OL]. (2015-08-26) [2022-12-13]. https://arxiv.org/abs/1508.06576.

[12] HeK, ZhangX, RenS, et al.. Deep residual learning for image recognition[C], 2016, New York, IEEE: 16541111

[13] SergeyK, MatthewT, HelenH, et al.. Recognizing image style[C], 2014, Durham, BMVA Press: 1-11

[14] Da SilvaD V C, RochaA A D A, VellosoP B. Mobile vs. non-mobile live-streaming: a comparative analysis of users engagement and interruption using big data from a large CDN perspective[J]. Sensors, 2021, 21: 5616

[15] FeiN, GaoY, LuZ, et al.. Z-score normalization, hubness, and few-shot learning[C], 2022, New York, IEEE

[16] LiG, ZhangM, LiJ, et al.. Efficient densely connected convolutional neural networks[J]. Pattern recognition, 2021, 109: 107610

[17] OtchereD A, GanatT O A, GholamiR, et al.. Application of supervised machine learning paradigms in the prediction of petroleum reservoir proper-ties: comparative analysis of ANN and SVM models[J]. Journal of petroleum science and engineering, 2021, 200: 108182

[18] HanS, KimH, LeeY S. Double random forest[J]. Mach learn, 2020, 109: 1569-1586