regions) Assumption: This dataset includes factors which are all related to the energy needs and consumption behaviour of a region Limitations: Not enough regions, too high level information (individual households would be much more interesting) and lacking many and possibly the most relevant characteristics (e.g. trafﬁc, population density, industry, connections between the variables)
the dataset, how many segments might be useful? E.g. elbow plot (k-means clustering, plotting SSE for possible number of segments) https://towardsdatascience.com/custo mer-segmentation-using-k-means-clu stering-d33964f238c3 3 groups
(e.g. regions with high number of one person households, newer buildings -> possibly high potential region? but what about the red dots?) • Finding proxies to energy consumption in each region and see how it relates to our segmentation (could be number of devices, check data from Swisscom) • Example question which could be answered: Is the assumed high potential region consuming less or more power? Example limitation: Since we do not have many regions and miss out characteristics, outlier (e.g. a region with a great number of unemployed persons) may end up in a not appropriate segment (e.g. maybe they live in regions with otherwise low potential and then therefore may receive no offer, but actually would be persons more willing to spend time to select a better provider and they are also the ones with greatest need to beneﬁt)
• Worldwide view: rnaturalearth is an R package to hold and facilitate interaction with natural earth vector map data. • National view: geofaceting provides a functionality for 'ggplot2'. Geofaceting arranges a sequence of plots of data for different geographical entities into a grid that preserves some of the geographical orientation.
in the world, from almost any device. This is dramatically changing the way people work, facilitating 24/7 collaboration with colleagues who are dispersed across time zones, countries, and continents.” Michael Dell, Chairman and CEO of Dell