Akademik Veri Yönetim Sistemi
IX.INTERNATIONAL HALICH CONGRESS ON MULTIDISCIPLINARY SCIENTIFIC RESEARCH, İstanbul, Türkiye, 3 - 04 Aralık 2025, cilt.1, ss.248-258, (Tam Metin Bildiri)
Although there are various communication channels, phone calls still play an important role in this process. However, spam calls pose significant risks and need to be addressed. If not detected, they can lead to financial and security issues, as well as a considerable loss of time. Therefore, spam call detection is crucial to mitigate these problems. Depending on the stage of the call, the spam call issue can be addressed in three phases: pre-ring, mid-call, and post-call. There are both familiar and distinct factors that can be used to detect spam calls at each phase of the time-based categorization. These factors play a key role in building an effective scoring mechanism. Therefore, this study examines them in detail to identify the most impactful ones, along with their importance weights, in order to provide a solid theoretical foundation for the scoring system. It has been identified that the effective factors and their corresponding importance weights vary across the different phases of a call. In the pre-ring phase, the most influential factors are the reputation of the caller (0.313), the blocklist status of the caller (0.275), the call intensity metrics (0.160), the social features of the caller (0.099), the call outcome metrics (0.088), the caller identity (0.038), and the call origin (0.026). During the midcall phase, the most essential factor groups/factors are voice activity (0.403), speech content (0.257), speaker identity (0.181), voice quality (0.079), acoustic features (0.051), and noise level (0.029). The post-call phase is only considered as the updating phase for factor weights. To summarize, the most crucial spam call factor for the pre-ring phase was determined as the reputation of the caller, with a weight of 0.41, and the least crucial pre-ring factor was the call origin. While voice activity became the most crucial spam call factor for the mid-call phase with a weight of 0.403, noise level took place at the end with a weight of 0.029. These findings provide a comprehensive basis for developing a phase-specific scoring mechanism for spam call detection.