Music recommender using deep embedding-based features and behavior-based reinforcement learning

Considering emotions and contextual factors in music recommendation: a systematic literature review

“…Mobile sensors or social networks could identify user behavior. Cheng and Shen [25] UH,CCI X 10 Helmholz et al [46] SEN, CCI X 11 Hong et al [49] UH, CCI, FEED X 12 Teng et al [117] MDI, UA, CCI, ML, CF X 13 Okada et al [96] ML, UB, MD X 14 Karlsson et al [67] UH, MDI X 15 Nirjon et al [95] SEN, UA, MDI X 16 Wang et al [126] MDI, UA, MD, ML X 17 Miller et al [88] CCI, MD X 18 Chang et al [21] MD, ML -19 Dias et al [32] UA, MD -20 Hansen et al [44] UH, CCI, ML -21 Wang et al [124] CF, MC, MD, UB -22 Assuncao and Neris [7] DIM, SI, MD Emotion X 23 Kang and Seo [66] DIM, MDI, CCI X 24 Abdul et al [2] DISC, SI, SIM, ML X 25 Chiu and Ko [26] DISC, SEN, ML, SI X 26 Iyer et al [57] SEN, FE X 27 Rosa et al [101] DIM, SN X 28 Guo et al [41] SEN, SIM X 29 Lehtiniemi and Holm [79] DIM, SI X 30 Helmholz et al [47] SI, DIM, DISC X 31 Andjelkovic et al [5] DIM, MD,SIM -32 Lopes et al [84] DISC, ML -33 Li and Liu [80] UB, MP -34 Padovani et al [97] DISC, SI, ML -35 Gilda et al [39] DISC, FE, MD ML -36 Vateekul et al [120] DIM, CF, SIM -37 Narducci et al [93] DISC, SN -…”

“…On the other hand, some studies have used machine learning techniques in music recommendation [61,78,21]. Jiang and He [61] incorporate an evolutionary model called Incremental Regression Tree, which collects contextual data, musical data, and user feedback in an incremental way to characterize his/her personal taste for music.…”

“…The recognition of user activities uses machine learning methods with smartphones signal resource selection techniques and collaborative filtering methods to estimate the user's musical preference. Other study that also incorporated machine learning methods for music recommendation was Chang et al [21]. The authors used reinforcement learning algorithms to learn the user's preferences and then make the recommendations.…”

“…Most current recommendation systems employ the collaborative filtering model, and these systems are vulnerable to the cold-start problem [21]. The cold-start problem arises because when a recommendation system has a new user, little information is known about that user.…”

“…The learning of MPlist is constituted of machine learning algorithms and implicit feedback to improve accuracy in forecasting. In part, Chang et al [21] used a machine learning method called booster learning, which makes a sequence of decisions to learn the user's preferences from the music playback record.…”