The Art of Personalisation: A Machine Learning Approach to Sex-Based Audiences in Mobile Advertising : Inference on Physical Gestures
Shadbolt, Jeremias (2024-06-03)
The Art of Personalisation: A Machine Learning Approach to Sex-Based Audiences in Mobile Advertising : Inference on Physical Gestures
(03.06.2024)
Julkaisu on tekijänoikeussäännösten alainen. Teosta voi lukea ja tulostaa henkilökohtaista käyttöä varten. Käyttö kaupallisiin tarkoituksiin on kielletty.
suljettu
Julkaisun pysyvä osoite on:
https://urn.fi/URN:NBN:fi-fe2024062056382
https://urn.fi/URN:NBN:fi-fe2024062056382
Tiivistelmä
In 2022, the global advertising technology market achieved a substantial valuation of USD 886.19 billion and is anticipated to sustain its growth momentum with a compound annual growth rate (CAGR) of 13.7% from 2023 to 2030. Ad Tech, encompassing technology and software for online advertisement management and measurement, has become a cornerstone of the digital marketing ecosystem.
Notably, the mobile segment dominated the Ad Tech market in 2022, commanding a significant 58.9% market share, and is poised for remarkable growth with a projected CAGR of over 14.0% during the forecast period of 2023 to 2030. Within this context, the role of targeting in marketing and advertising assumes paramount importance. Previous research underscores the effectiveness of Behavioural Targeting (BT), which leads to significant increase in conversion rates and revenue. This underscores the importance of proper targeting as the market continues to grow.
However, this thriving Ad Tech landscape faces formidable challenges. Leading mobile industry players such as Google and Apple are increasing focus in prioritizing security and user privacy, inherently hindering previous targeting solutions. This introduces challenges to creating reliable audiences whilst ensuring targeting accuracy. In parallel, user awareness of data privacy issues, both within the Ad Tech sphere and in general, is on the rise. This highlights a pressing need for innovative approaches to crafting privacy-focused cohorts while maintaining accuracy, despite the apparent tension.
Single-touch and multi-touch gestures have emerged as the dominant means of engaging with technology, including smartphones, tablets, and various other touchscreen devices. Recent advancements in touchscreen input capabilities have substantially augmented the volume and quality of touch gesture data, prompting its investigation in diverse fields ranging from psychology to biometrics.
This thesis aimed to research how to create privacy-focused still accurate sex-based audiences by utilising touch gestures on mobile devices. It is built on a literature review that studies the current state of Ad Tech in the context of mobile advertising and gesture-based user segmentation. Then, an empiric research, based on current literature, is conducted with the data collected through a custom mobile application utilised to gather a dataset of gestures, as other datasets used for similar research proved challenging to obtain.
This thesis is a part of a broader internal gesture-based project at Verve Group and Media and Games Invest SE.
The results of the research and the literature review were that men exert longer and faster swipes with larger physical movement of the device, indicating discriminating differences thus opening a possibility for machine learning. A novel approach in training was introduced through swipe aggregation, as well as three new features not utilised in academia. In addition, a deep learning approach was introduced as deep learning had not been used in previous research.
Notably, the mobile segment dominated the Ad Tech market in 2022, commanding a significant 58.9% market share, and is poised for remarkable growth with a projected CAGR of over 14.0% during the forecast period of 2023 to 2030. Within this context, the role of targeting in marketing and advertising assumes paramount importance. Previous research underscores the effectiveness of Behavioural Targeting (BT), which leads to significant increase in conversion rates and revenue. This underscores the importance of proper targeting as the market continues to grow.
However, this thriving Ad Tech landscape faces formidable challenges. Leading mobile industry players such as Google and Apple are increasing focus in prioritizing security and user privacy, inherently hindering previous targeting solutions. This introduces challenges to creating reliable audiences whilst ensuring targeting accuracy. In parallel, user awareness of data privacy issues, both within the Ad Tech sphere and in general, is on the rise. This highlights a pressing need for innovative approaches to crafting privacy-focused cohorts while maintaining accuracy, despite the apparent tension.
Single-touch and multi-touch gestures have emerged as the dominant means of engaging with technology, including smartphones, tablets, and various other touchscreen devices. Recent advancements in touchscreen input capabilities have substantially augmented the volume and quality of touch gesture data, prompting its investigation in diverse fields ranging from psychology to biometrics.
This thesis aimed to research how to create privacy-focused still accurate sex-based audiences by utilising touch gestures on mobile devices. It is built on a literature review that studies the current state of Ad Tech in the context of mobile advertising and gesture-based user segmentation. Then, an empiric research, based on current literature, is conducted with the data collected through a custom mobile application utilised to gather a dataset of gestures, as other datasets used for similar research proved challenging to obtain.
This thesis is a part of a broader internal gesture-based project at Verve Group and Media and Games Invest SE.
The results of the research and the literature review were that men exert longer and faster swipes with larger physical movement of the device, indicating discriminating differences thus opening a possibility for machine learning. A novel approach in training was introduced through swipe aggregation, as well as three new features not utilised in academia. In addition, a deep learning approach was introduced as deep learning had not been used in previous research.