paulqw / iot-analytics

Market basket analysis

Date: 02/09/2019

Project summary

Business case: Working as an IoT-Analyst at the fictitious company IoT-Analytics for a developer(client) of Smart Homes. Their major goal is to develop highly efficient Smart Homes make use of cutting-edge technology like electrical sub-metering devices and power management enhanced by IoT-Analytics.
To support the client and empower Smart Home owners with greater understanding and control of their electrical power usage a large dataset has been analysed containing 47 month of energy usage data of an average house.

Goal of analysis: The goal of analysis is to perform time series analysis to discover seasonalities and trends within a typical profile of a family’s electrical consumption. Based on the analysis meaningful visualization strategy has been pursued to add the most value to potential customers of the client - Smart Home owners. Hence, both the total electrical power consumption and available data for submetering-devices installed for the laundry, kitchen, and heating/AC appliances have been analysed. Mainly, the analysis unveils the total yearly amount of energy consumed, allows a deep dive in the consumption split by submeters, seeks to forecast the electrical consumption, and demonstrates possible saving potential. Finally, results are visualized and organised in a dashboard using Power BI.

Technical approach of analysis

1. Exploration and preparation of data

• Set time-zone, format, and intervals
• Identify, pad, and fill NAs
• Aggregate and check different granularity

2. Stationarity, autocorrelation, and transformation

• Create and analysis (multi-seasonal) decomposition
• Check stationarity and autocorrelation
• Ensure stationarity/conduct transformations

3. (Multi-seasonal) Time series and data splitting

• Create time series and split data

4. Train models, evaluate, forecast

• Train several forecasting models (e.g. ARIMA, prophet)
• Perform cross-validation
• Evaluate models based on several performance measures such as RMSE, MAE, MAPE
• Forecast energy consumption

5. Create visualization and dashboard