We study human behavior from a Complex Systems and Computational Social Science approach within the PICAE project (Intelligent Publication of Audiovisual and Editorial Contents), where we are studying the TV and consumption behavior from the Complex Systems Science perspective. The aim is to characterize behavioral patterns and eventually use this knowledge to improve the performance of the audiovisual content recommender that another partner from PICAE will be building. The data cover the last 3 years (2019-2021) of TV consumption (CCMA, Corporació Catalana de Mitjans Audiovisuals, TV3), hence we are dealing with the order of millions of data (~2-3 millions), which is carefully analyzed, processed and cleaned. In the statistical study of the data, we observed three well-known audience peaks (early morning, afternoon and evening), as well as characteristic durations of the video clips consumed. The duration of consumption does not have a characteristic time scale and is very varied. We also observed differences in consumption depending on the time of year and the day of the week, as well as on the device used. We use a mixed data clustering algorithm to group the consumption units into 5 distinct groups, which we can basically characterized by the time of consumption, the durations of video clips and consumption and the type of device used, as the most relevant variables. Using a machine learning algorithm we can train and predict the clusters found with high accuracy, and therefore we can determine the importance or impact of the consumption variables in each of the clusters.