Supplementary Dataset for: Resource Demand Prediction for Network Slices in 5G using ML Enhanced with Network Models

The original MILANO dataset (https://ieee-dataport.org/documents/milan-dataset) examines the relationship between meteorological context and cellular traffic loads, telecommunication and weather data from the city of Milan is presented. The dataset consists of aggregated telecommunication and weather data from the city of Milan during the period of 1st of November 2013 to 1st of January 2014. The telecommunication data consists of aggregated information of received SMS, sent SMS, incoming call, outgoing call, and internet activity, and is measured through Call Detail Records (CDRs), a measure of volume of cellular traffic. The weather data consists of information on humidity, measured in percentage, temperature, measured in degrees Celsius, and wind speed, measured in meters per second. The data is aggregated by hours and is listed over an area of 300 meters squared within the city. The city is divided into grids of 1x1m, and a binary adjacency matrix is presented to represent the connectivity between the grids in the city.Within the paper “Resource Demand Prediction for Network Slices in 5G Using ML Enhanced With Network Models”, a post-processed version of the MILANO dataset was created, which includes the following attitional fields, to facilitate the ability to forecast the slice resource demand. The following fields are included:

1.    bsId: The 5G BaseStation ID. We can have multiple associations between the same BaseStation ID, like several occurrences of the same ID that plays the role of Input Cell and Output transmission Cell.2.    Episode. The game-theoretic episode index of the decentralized AI consensus admission control algorithm(s) utilized.3.    Step. Simply a numerical index starting from 0, and augmenting.4.    loadSMS. Total in/out SMS traffic load.5.    RawActSMS. Original raw network packets (SMS load).6.    loadInt. Total in/out Internet traffic load.7.    rawActInt. Original raw network packets (Internet load).8.    loadCalls. Total in/out Voice traffic load.9.    rawActCalls. Original raw network packets (Voice load).10.    defSMS. Deferred Total in/out SMS traffic load.11.    rawActDefSMS. Deferred Original raw network packets (SMS load).12.    defInt. Deferred Total in/out Internet traffic load.13.    rawActDefInt. Deferred Original raw network packets (Internet load).14.    defCalls. Deferred Total in/out Voice traffic load.15.    rawActDefCalls. Deferred Original raw network packets (Voice load).16.    schedSMS. Admitted SMS traffic load.17.    schedInt. Admitted Internet traffic load.18.    schedCalls. Admitted Voice Traffic load.19.    schedDelaySMS. Queued for admission SMS traffic load.20.    schedDelayInt. Queued for admission Internet traffic load.21.    schedDelayCalls. Queued for admission Voice Traffic load.22.    dropSMS. Raw Packets Dropped SMS traffic load.23.    dropInt. Raw Packets Dropped Internet traffic load.24.    dropCalls. Raw Packets Dropped Voice Traffic load.25.    totalSched. Total Admitted Traffic load.26.    bsCap. BaseStation network capacity.27.    totalDropped. Total Raw Packets Dropped Traffic load.28.    rejectRate. Total Traffic load network packets reject rate.29.    rejectRateCalls. Voice Traffic network packets reject rate.30.    rejectRateInt. Internet traffic network packets reject rate.31.    rejectRateSMS. SMS traffic network packets reject rate.32.    delayRate. Total Delayed (not queued yet for admission) Traffic load.33.    delayRateCalls. Total Delayed (not queued yet for admission) Voice Traffic load.34.    delayRateInt. Total Delayed (not queued yet for admission) Internet traffic load.35.    delayRateSMS. Total Delayed (not queued yet for admission) SMS traffic load.36.    Reward. Total game reward value.37.    episodeReward. Total reward value per episode.38.    avgEpisodeReward. Average reward value per episode.