Evaluation of Machine Leaning Algorithms for Streets Traffic Prediction: A Smart Home Use Case

Area of Science:

• Computer Science
• Artificial Intelligence
• Energy Systems

Background:

• Smart homes aim to reduce energy consumption.
• Current methods often focus on in-home behavior prediction.
• Optimizing heating, cooling, and hot water systems is key to energy efficiency.

Purpose of the Study:

• To define a smart home use case for automatic temperature and hot water adjustment.
• To reduce energy consumption in smart homes by optimizing climate control systems.
• To identify the most effective machine learning and neural network (MLNN) algorithms for predicting driver arrival times.

Main Methods:

• A machine learning-based street traffic prediction system is employed.
• Focuses on predicting resident arrival times outside the home.
• Evaluates various MLNN algorithms using Uber's San Francisco dataset and imputation for missing values.

Main Results:

• The developed system accurately predicts driver arrival times.
• Identifies optimal MLNN algorithms for traffic prediction.
• The system can also function as a route recommender.

Conclusions:

• Predicting external resident behavior, like arrival time, is effective for smart home energy reduction.
• Machine learning algorithms can significantly enhance smart home energy management.
• The traffic prediction system offers dual benefits of energy savings and route optimization.