Abstract:Have you ever considered the sheer volume of actions we perform using IoT (Internet of Things) devices within our homes, offices, and daily environments? From the mundane act of flicking a light switch to the precise adjustment of room temperatures, we are surrounded by a wealth of data, each representing a glimpse into user behaviour. While existing research has sought to decipher user behaviours from these interactions and their timestamps, a critical dimension still needs to be explored: the timing of these actions. Despite extensive efforts to understand and forecast user behaviours, the temporal dimension of these interactions has received scant attention. However, the timing of actions holds profound implications for user experience, efficiency, and overall satisfaction with intelligent systems. In our paper, we venture into the less-explored realm of human-centric AI by endeavoring to predict user actions and their timing. To achieve this, we contribute a meticulously synthesized dataset comprising 11k sequences of actions paired with their respective date and time stamps. Building upon this dataset, we propose our model, which employs advanced machine learning techniques for k-class classification over time intervals within a day. To the best of our knowledge, this is the first attempt at time prediction for smart homes. We achieve a 40% (96-class) accuracy across all datasets and an 80% (8-class) accuracy on the dataset containing exact timestamps, showcasing the efficacy of our approach in predicting the temporal dynamics of user actions within smart environments.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to predict the time of the user's next action in the smart home environment. Although existing research has been able to decode user behavior from the interaction between users and Internet - of - Things devices, these studies mainly focus on predicting what the user will do next, ignoring the specific time when these behaviors occur. However, for improving user experience, efficiency, and overall satisfaction with the intelligent system, the timing of the behavior is equally important. Therefore, the authors of this paper propose a new method to predict the user's behavior and its occurrence time in the smart home environment, aiming to fill this research gap. Specifically, the authors constructed a data set containing 11,000 action sequences, each with an accurate timestamp, and proposed a model that uses advanced machine - learning techniques to perform multi - classification (k - class classification) on time intervals within a day. Through this method, the authors can not only predict the actions that the user may perform next, but also predict when these actions will occur, thus providing a more personalized and proactive service experience for the user. For example, if the system can predict that the user usually starts preparing breakfast at 7:30 in the morning, it can pre - heat the oven or prepare the coffee machine in advance, providing a seamless and proactive user experience. The significance of this research lies in that by introducing the dimension of time prediction, the responsiveness and effectiveness of the intelligent system can be significantly improved, enabling the intelligent system to support users more proactively rather than merely passively responding to user commands. This marks the first step towards creating an automated - assistance AI system that can significantly improve the user experience in smart homes.

Timing Matters: Enhancing User Experience through Temporal Prediction in Smart Homes

On Using Temporal Features to Create More Accurate Human-Activity Classifiers

I Know Your Intent

Effective Anomaly Detection in Smart Home by Integrating Event Time Intervals

Modeling Interdependent and Periodic Real-World Action Sequences

Learning Activity Predictors from Sensor Data: Algorithms, Evaluation, and Applications

A human-centered smart home system with wearable-sensor behavior analysis

Potential Impacts of Smart Homes on Human Behavior: A Reinforcement Learning Approach

Room Occupancy Prediction: Exploring the Power of Machine Learning and Temporal Insights

User Device Interaction Prediction Via Relational Gated Graph Attention Network and Intent-aware Encoder

A Novel Human Activity Recognition and Prediction in Smart Home Based on Interaction

The House That Knows You: User Authentication Based on IoT Data

Towards Detecting IoT Event Spoofing Attacks Using Time-Series Classification

Deep learning based predictive analysis of energy consumption for smart homes

IoT-Based Household Energy Consumption Prediction Using Machine Learning

A novel user behavior prediction model based on automatic annotated behavior recognition in smart home systems

Behavior Prediction Using an Improved Hidden Markov Model to Support People with Disabilities in Smart Homes

Anticipating Daily Intention using On-Wrist Motion Triggered Sensing

Human Action Anticipation: A Survey

Jointly Prediction of Activities, Locations, and Starting Times for Isolated Elderly People

Occupancy Prediction in IoT-Enabled Smart Buildings: Technologies, Methods, and Future Directions