In the case of a real smart home, if the results revealed the need to change something, this is usually a costly and infeasible choice to make. The results of the evaluation could reveal the need to redesign the smart home by adding or removing smart devices or changing the scenarios generated, etc.

Then, the model will be tested on another part of the same dataset, and its results will be evaluated. The model creation starts by initialising its parameters and training on a portion of the dataset. A common practice in machine learning is to divide the dataset into two parts, training and testing. When developing machine learning models, targeting specific functionalities, researchers rely on the existence of good representative datasets. Additionally, they can offer solutions such as the ability to pause and fast-forward the simulation to enable more accurate activity annotation. Such tools facilitate fast dataset generation and offer robust methods to capture the sensors’ data. The existence of a dataset simulation tool overcomes the drawbacks/challenges of generating real datasets. We conducted a study to assess the ease of use of our tool on the System Usability Scale (SUS). Our tool divides the dataset generation process into three distinct phases: first design: the researcher designs the initial virtual environment by building the home, importing smart devices and creating contexts second, simulation: the participant simulates his/her context-specific events and third, aggregation: the researcher applies the replication algorithm to generate the final dataset. We have built an extensible library of smart devices that facilitates the simulation of current and future smart home environments. The replication provides a solution for generating large representative smart home datasets. A small sample dataset produced, by OpenSHS, can be extended without affecting the logical order of the events. We have designed a replication algorithm for extending and expanding a dataset. This approach reduces the time and efforts required to generate simulated smart home datasets. Following a hybrid approach, OpenSHS combines advantages from both interactive and model-based approaches.

OpenSHS offers an opportunity for researchers in the field of the Internet of Things (IoT) and machine learning to test and evaluate their models. This paper develops a new hybrid, open-source, cross-platform 3D smart home simulator, OpenSHS, for dataset generation.

Jmonkeyengine