Scientific Programming

Research Article

A High-Frequency Data-Driven Machine Learning Approach for Demand Forecasting in Smart Cities

Table 1

Related work comparison with respect to frequency, forecast horizon, and other factors or complex estimation needed to apply the approach.
WorkRelated work comparison
FrequencyForecast horizonOther factors
Adamowski et al. [20]1 for each day24 hoursWeather data during summer
Herrera et al. [21]1 for each hour1 hourWeather data
Odan and Reis [22]1 for each hour24 hoursWeather data
Ji et al. [23]1 for each hour24 hoursWeather, holidays, and incident data
Hutton and Kapelan [24]1 for each hour24 hoursAnnual calendar data
Candelieri et al. [25–27]1 for each hour24 hoursWorking days and seasons of the year
Alvisi and Franchini [28]1 for each hour24 hoursWeekly seasonality and seasons of the year
Brentan et al. [29]1 for each hour24 hoursAnnual calendar data
Romano and Kapelan [30]1 for each hour24 hoursWeekly seasonality
Gagliardi et al. [31]1 for each hour24 hoursWeekly seasonality
Pacchin et al. [32]1 for each hour24 hoursWeekly seasonality
Arandia et al. [33]1 for each 15 minutes24 hoursDaily and weekly seasonality
Bakker et al. [34]1 for each 15 minutes48 hoursAnnual calendar data
Our proposal1 for each minute24 hoursWeekly seasonality