Estimating the number of household TV profiles based in customer behaviour using Gaussian mixture model averaging

TL;DR

This work tackles the problem of estimating how many household TV profiles exist within a viewing dataset where explicit per-person labels are unavailable. It introduces a novel framework that couples Gaussian Mixture Model averaging to produce point estimates $G_{est}$ of the number of profiles with a Bayesian random-walk model to quantify temporal uncertainty. Using BT YouView data (from $228$ customers and roughly $5\times 10^5$ observations) and 17 engineered features, the authors show that dimensionality reduction to four latent variables explains $65.9\%$ of the variance and increases the estimated profile count to $G_{est}$ values around $9.16$ (sd $3.93$) versus $4.87$ (sd $3.96$) when using all features. The Bayesian component provides time-varying credible intervals, enabling uncertainty-aware recommendations and planning, and lays groundwork for integrating these profiles into future recommendation engines for both individual and group viewing. Overall, the approach offers a data-driven path to improve personalised content delivery and targeted advertising in multi-user TV environments, with clear avenues for extending to daily estimates and operational integration.

Abstract

TV customers today face many choices from many live channels and on-demand services. Providing a personalised experience that saves customers time when discovering content is essential for TV providers. However, a reliable understanding of their behaviour and preferences is key. When creating personalised recommendations for TV, the biggest challenge is understanding viewing behaviour within households when multiple people are watching. The objective is to detect and combine individual profiles to make better-personalised recommendations for group viewing. Our challenge is that we have little explicit information about who is watching the devices at any time (individuals or groups). Also, we do not have a way to combine more than one individual profile to make better recommendations for group viewing. We propose a novel framework using a Gaussian mixture model averaging to obtain point estimates for the number of household TV profiles and a Bayesian random walk model to introduce uncertainty. We applied our approach using data from real customers whose TV-watching data totalled approximately half a million observations. Our results indicate that combining our framework with the selected features provides a means to estimate the number of household TV profiles and their characteristics, including shifts over time and quantification of uncertainty.

Figures (4)

• Figure 1: A diagram illustrating the complete proposed approach to estimate the number of household TV profiles and providing recommendations based on the following number.
• Figure 2: Scatter plot of point estimates of household TV profiles using Gaussian Mixture Models averaging and the ratio of average distance within and between clusters obtained from the Exploratory Factor Analysis latent variables and all features of watching behaviour.
• Figure 3: Cumulative density function of the estimated household TV profiles over ten months (December 2020 to September 2021) for all customers included in our dataset.
• Figure 4: Time series of household TV profiles over ten months (December 2020 to September 2021) for nine customers, where lines represent the posterior means alongside the 95% credible intervals as the gray ribbons. The points represent the point estimates provided by the Gaussian mixture model averaging.