Serendipity in recommender systems (RSs) has attracted increasing attention as a concept that enhances user satisfaction by presenting unexpected and useful items. However, evaluating serendipitous performance remains challenging because its ground truth is generally unobservable. The existing offline metrics often depend on ambiguous definitions or are tailored to specific datasets and RSs, thereby limiting their generalizability. To address this issue, we propose a universally applicable evaluation framework that leverages large language models (LLMs) known for their extensive knowledge and reasoning capabilities, as evaluators. First, to improve the evaluation performance of the proposed framework, we assessed the serendipity prediction accuracy of LLMs using four different prompt strategies on a dataset containing user-annotated serendipitous ground truth and found that the chain-of-thought prompt achieved the highest accuracy. Next, we re-evaluated the serendipitous performance of both serendipity-oriented and general RSs using the proposed framework on three commonly used real-world datasets, without the ground truth. The results indicated that there was no serendipity-oriented RS that consistently outperformed across all datasets, and even a general RS sometimes achieved higher performance than the serendipity-oriented RS.

Urban air pollution is a major environmental problem affecting human health and quality of life. Monitoring stations have been established to obtain air quality information continuously, but they do not cover all areas. Thus, there are numerous methods for spatially fine-grained air quality inference. Since existing methods aim to infer air quality of locations only in monitored cities, they do not assume inferring air quality in unmonitored cities. In this paper, we first study the air quality inference in unmonitored cities. To accurately infer air quality in unmonitored cities, we propose a neural network-based approach AIREX. The novelty of AIREX is employing a mixture-of-experts approach, a machine learning technique based on the divide-and-conquer principle, to learn correlations of air quality between multiple cities. To further boost the performance, it employs attention mechanisms to compute the impacts of air quality inference from the monitored cities to the locations in the unmonitored city. Through experiments on a real-world air quality dataset, we show that AIREX achieves higher accuracy than state-of-the-art methods.

Abstract This article addresses a new pattern mining problem in time series sensor data, which we call correlated attribute pattern mining. The correlated attribute patterns (CAPs for short) are the sets of attributes (e.g., temperature and traffic volume) on sensors that are spatially close to each other and temporally correlated in their measurements. Although the CAPs are useful to accurately analyze and understand spatio-temporal correlation between attributes, the existing mining methods are inefficient to discover CAPs because they extract unnecessary patterns. Therefore, we propose a mining method Miscela to efficiently discover CAPs. Miscela can discover not only simultaneous correlated patterns but also time delayed correlated patterns. Furthermore, we extend Miscela to automatically search for correlated patterns with any time delays. Through our experiments using three real sensor datasets, we show that the response time of Miscela is up to 20.84 times faster compared with the state-of-the-art method. We show that Miscela discovers meaningful patterns for urban managements and environmental studies.

Estimating land prices is useful for assessing values of sites. Several works study estimating land prices from land features that are extracted from geodetic data. However, the estimation accuracy is not high enough yet because it is difficult to thoroughly collect geodetic data that affects land prices. In this paper, we study the effectiveness of the satellite images to estimate land prices for the first time. To verify effectiveness of satellite images, we estimate land prices by using three deep neural network models: multilayer perceptrons (MLP) model only with geodetic data, convolution neural network (CNN) model only with satellite images, and concatenation model that concatenates the MLP with the CNN models. We demonstrate through experiments using real land prices, geodetic data, and satellite images in Japan that the simultaneous use of satellite images and geodetic data improves the estimation accuracy of land prices.

The urban condition is monitored by a wide variety of sensors with several attributes such as temperature and traffic volume. It is expected to discover the correlated attributes to accurately analyze and understand the urban condition. Several mining techniques for spatio-temporal data have been proposed for discovering the sets of sensors that are spatially close to each other and temporally correlated in their measurements. However, they cannot discover correlated attributes efficiently because their targets are correlated sensors with a single attribute. In this paper, we introduce a problem of discovering correlations among multiple attributes, which we call correlated attribute pattern (CAP) mining. Although the existing spatio-temporal data mining methods can be extended to discover CAPs, they are inefficient because they extract unnecessary correlated sensors that do not have CAPs. Therefore, we propose a CAP mining method MISCELA to efficiently discover CAPs. In MISCELA, we develop a new tree structure called CAP search tree, by which we can effectively prune the unnecessary patterns for the CAP mining. Our experiments using real sensor datasets show that the response time of MISCELA is up to 79% faster compared to the state-of-the-art.

This paper proposed the method to reduce the calculating time to reveal the functional brain network associated with a task using a genetic algorithm and functional near-infrared spectroscopy (fNIRS) data. Changes in the cerebral blood flow during a task are obtained as time series data is analyzed using fNIRS, and a correlation matrix for multiple fNIRS channels is created for each subject. The subject group is divided into two groups, and a classifier of the two groups learns the correlation matrix as a feature quantity. The correlation matrix changes as the feature quantity changes with the combinations of channels, which affects classifier accuracy. If the combination of channels with the best classifier accuracy is identified, these channels can be considered important to the creation of the functional brain network for a target task. In our study, a genetic algorithm (GA) is used for channel selection. However, learning the classifier to calculate the evaluation value and optimization by the GA requires significant time. Thus, to increase search efficiency, we propose the kick-out method to skip the evaluation value calculation for poor individuals according to a previous evaluation value. We evaluated the effectiveness of the proposed method using fNIRS data recorded during a mental rotation test. Results show that important channels that express the functional brain network were selected and that processing time was reduced significantly by the proposed method.