Accurate forecasting of residential energy consumption is essential for improving energy efficiency, lowering costs, and enabling effective grid management. This study examines how varying prediction interval resolutions impact the accuracy and execution time of nine machine learning models in the context of household energy use. Using an existing dataset of household energy consumption containing over 2 million records collected from 2006 to 2010, we downsample the time-series data into multiple intervals to assess how changing time periods influence prediction performance. Nine machine learning models are trained and tested across these different resolutions in order to identify how temporal granularity affects predictive reliability. Our results suggest that high-resolution data may not always be required since some models show stable performance despite reduced data frequency.

Formal specification languages have a lot of notions in common. They all introduce entities usually called processes, offer similar operators, and most importantly define their operational semantics based on labeled transition systems (LTS). However, each language defines specific synchronizing and/or memory structures. For instance, in CSP, the synchronization is defined between identical events, while in CCS and in synchronization vectors-based views it is defined respectively between complementary events or between possibly different events. In this paper, we aim at capturing some similarities of specification languages by defining a label-based composition formal framework. Firstly, we define a high-level synchronization mechanism in the form of an abstract label structure. We then couple this label structure with several compositional operations and properties. Secondly, we introduce an LTS-based behavioral framework and define a unique LTS composition operator which is reused to define syntactic composition of extended transition systems and a compositional semantics.