Time Series Clustering Algorithm Research Articles

A dynamic customer segmentation approach by combining LRFMS and multivariate time series clustering

To successfully market to automotive parts customers in the Industrial Internet era, parts agents need to perform effective customer analysis and management. Dynamic customer segmentation is an effective analytical tool that helps parts agents identify different customer groups. RFM model and time series clustering algorithms are commonly used analytical methods in dynamic customer segmentation. The original RFM model suffers from the problems of R index randomness and ignoring customers’ perceived value. For most existing studies on dynamic customer segmentation, time series clustering techniques largely focus on univariate clustering, with less research on multivariate clustering. To solve the above problems, this paper proposes a dynamic customer segmentation approach by combining LRFMS and multivariate time series clustering. Firstly, this method represents each customer behavior as a time series sequence of the Length, Recency, Frequency, Monetary and Satisfaction variables. And then, we apply a multi-dimensional time series clustering algorithm based on three distance measurement methods called DTW-D, SBD, and CID to carry out customer segmentation. Finally, an empirical study and comparative analyses are conducted using customer transaction data of parts agents to verify the effectiveness of the approach. Additionally, a detailed analysis of different customer groups is made, and corresponding marketing suggestions are provided.

Time series clustering based on polynomial fitting and multi-order trend features

Time series clustering serves as a potent data mining method, facilitating the analysis of an extensive array of time series data without the prerequisite of any prior knowledge. It finds wide-ranging use across various sectors, including but not limited to, financial and medical data analysis, and sensor data processing. Given the high dimensionality, non-linearity, and redundancy characteristics associated with time series, conventional clustering algorithms frequently fall short in yielding satisfactory results when directly applied to this kind of data. As such, there is a critical need to judiciously select suitable feature extraction methods and dimension reduction techniques. This paper introduces a time series clustering algorithm, drawing primarily from polynomial fitting derivative features as a wellspring for feature extraction to achieve effective clustering results. Initially, Hodrick Prescott (HP) filtering comes into play for the processing of raw time series data, thereby eliminating noise and redundancy. Subsequently, polynomial curve fitting (PCF) is applied to the data to derive a globally continuous function fitting this time series. Next, by securing multi-order derivative values via this function, the time series is transformed into a multi-order derivative feature sequence. Lastly, we designed a polynomial function derivative features-based dynamic time warping (PFD_DTW) algorithm for determining the distance between two equal or unequal granular length time series, and subsequently a hierarchical clustering method anchored on the PFD_DTW distances for time series clustering after computing interspecies distances. The effectiveness of this method is corroborated by experimental results obtained from several practical datasets.

Clustering time series by extremal dependence

AbstractThe goal of this paper is to characterize the temporal dependence structure on the extremes of time series and use such dependency to group them. In particular, three similarity measures to capture extremal dependence are proposed, being their performance assessed in different scenarios. This will involve the use of classical time series clustering algorithms, as well as rigorous evaluation of their performance in both simulated scenarios and real-world time series data sets. The focus will be on comparing the performance of these similarity measures with different clustering methods, and illustrate the efficacy of extremal dependence-based clustering in meteorological data. To achieve this, we will consider a dataset consisting of daily maximum temperatures recorded at 500 stations across Europe.

Construction of Cloud Computing Resource Allocation and Ideological and Political Education Platform Based on the Theme of Party History and Party Building

Abstract Party history and party-building education is an important carrier of party-building work in colleges and universities, and it is also an important content of cultivating and promoting socialist core values while party-building work is an effective way and the main hand of ideological and political work of students in colleges and universities. This paper takes the large amount of data generated in the process of ideological and political education resource allocation on the theme of party history and party building as the research object and builds a mapping relationship between tasks and ideological and political education resources according to certain strategies based on cloud computing resource allocation. A platform for allocating resources for ideological and political education has been designed and developed, which is based on a time series clustering algorithm and constructs resource overviews, sequence collections, and clustering modules. The functions of student behavior data collection and overview, feature calculation and selection, behavior time series clustering, and group feature analysis are realized, which enhances the efficiency of the ideological and political education resource platform. Functional and performance tests of the platform were completed, and the average system response times of the three modules were 75.05, 80.39, and 69.95ms, which were able to meet the actual application standards. With the help of the platform, the behavioral record data of 6,927 platform users were evaluated and practiced, and the user group classification and group characterization were carried out according to the output of the platform, which provided scientific suggestions for the distribution of ideological and political teaching resource services.

Time series clustering to improve one-class classifier performance

The improvement of one-class classifiers’ performance through clustering of multivariate time series is considered in this paper. Datasets arising from real processes come from the available sensors and are affected by many factors, such as aging of the process, changes in the operation region, and equipment malfunction. Despite that, one expects that the classes represented by such diverse data can be unveiled via trained classifiers. This work hypothesizes that the overall performance can be improved by training sets of one-class classifiers with subsets of data clustered by similarity. The proposed method is applied to one class classifiers since they are trained only with the target class, which is clustered based on time series similarity using Dynamic Time Warping and k-means. The advantages of the techniques are illustrated through their application to a public dataset from the oil industry with instances characterizing eight classes of data represented by five time series. Seven classes are selected to train LSTM classifiers using the variables and instances clustered using time series clustering algorithms. The results show that the increase in the similarity of training data tends to improve the performance of the LSTM classifier, achieving an increase of 10% in the overall performance. In a specific case, where the clustering model raised the similarity by 84%, the classification performance improved by 21%.

An air combat maneuver pattern extraction based on time series segmentation and clustering analysis

Target maneuver recognition is a prerequisite for air combat situation awareness, trajectory prediction, threat assessment and maneuver decision. To get rid of the dependence of the current target maneuver recognition method on empirical criteria and sample data, and automatically and adaptively complete the task of extracting the target maneuver pattern, in this paper, an air combat maneuver pattern extraction based on time series segmentation and clustering analysis is proposed by combining autoencoder, G-G clustering algorithm and the selective ensemble clustering analysis algorithm. Firstly, the autoencoder is used to extract key features of maneuvering trajectory to remove the impacts of redundant variables and reduce the data dimension; Then, taking the time information into account, the segmentation of Maneuver characteristic time series is realized with the improved FSTS-AEGG algorithm, and a large number of maneuver primitives are extracted; Finally, the maneuver primitives are grouped into some categories by using the selective ensemble multiple time series clustering algorithm, which can prove that each class represents a maneuver action. The maneuver pattern extraction method is applied to small scale air combat trajectory and can recognize and correctly partition at least 71.3% of maneuver actions，indicating that the method is effective and satisfies the requirements for engineering accuracy. In addition, this method can provide data support for various target maneuvering recognition methods proposed in the literature, greatly reduce the workload and improve the recognition accuracy.

Patterns and drivers of heat production in the plant genus Amorphophallus.

Thermogenesis - the ability to generate metabolic heat - is much more common in animals than in plants, but it has been documented in several plant families, most prominently the Araceae. Metabolic heat is produced in floral organs during the flowering time (anthesis), with the hypothesised primary functions being to increase scent volatilisation for pollinator attraction, and/or to provide a heat reward for invertebrate pollinators. Despite in-depth studies on the thermogenesis of single species, no attempts have yet been made to examine plant thermogenesis across an entire clade. Here, we apply time-series clustering algorithms to 119 measurements of the full thermogenic patterns in inflorescences of 80 Amorphophallus species. We infer a new time-calibrated phylogeny of this genus and use phylogenetic comparative methods to investigate the evolutionary determinants of thermogenesis. We find striking phenotypic variation across the phylogeny, with heat production in multiple clades reaching up to 15°C, and in one case 21.7°C above ambient temperature. Our results show that the thermogenic capacity is phylogenetically conserved and is also associated with inflorescence thickness. Our study paves the way for further investigations of the eco-evolutionary benefits of thermogenesis in plants.

Establishing zone regions in cyclic voltammetry using unsupervised machine learning

Cyclic voltammetry is widely used to establish mechanistic details associated with electrode processes. Commonly, an experienced electrochemist manually seeks to identify key feature maps to create zones of parameters present in simulated cyclic voltammograms that mimic features observed experimentally over a range of scan rates and concentrations. In this study, as an alternative to this tedious approach, an automated K-means time series clustering algorithm is applied as a form of unsupervised machine learning to sort and classify voltammetric information. In order to establish advantages and limitations of the K-means clustering approach, extensive simulations of six commonly encountered electrochemical reaction mechanisms using a wide range of parameters have been undertaken to generate zone regions relevant to each example. The K-means time series clustering together with a dynamic time warping metric successfully sort the simulated data into logical groups based on the features present.

Comparison of time series clustering methods for identifying novel subphenotypes of patients with infection.

Severe infection can lead to organ dysfunction and sepsis. Identifying subphenotypes of infected patients is essential for personalized management. It is unknown how different time series clustering algorithms compare in identifying these subphenotypes. Patients with suspected infection admitted between 2014 and 2019 to 4 hospitals in Emory healthcare were included, split into separate training and validation cohorts. Dynamic time warping (DTW) was applied to vital signs from the first 8h of hospitalization, and hierarchical clustering (DTW-HC) and partition around medoids (DTW-PAM) were used to cluster patients into subphenotypes. DTW-HC, DTW-PAM, and a previously published group-based trajectory model (GBTM) were evaluated for agreement in subphenotype clusters, trajectory patterns, and subphenotype associations with clinical outcomes and treatment responses. There were 12 473 patients in training and 8256 patients in validation cohorts. DTW-HC, DTW-PAM, and GBTM models resulted in 4 consistent vitals trajectory patterns with significant agreement in clustering (71-80% agreement, P < .001): group A was hyperthermic, tachycardic, tachypneic, and hypotensive. Group B was hyperthermic, tachycardic, tachypneic, and hypertensive. Groups C and D had lower temperatures, heart rates, and respiratory rates, with group C normotensive and group D hypotensive. Group A had higher odds ratio of 30-day inpatient mortality (P < .01) and group D had significant mortality benefit from balanced crystalloids compared to saline (P < .01) in all 3 models. DTW- and GBTM-based clustering algorithms applied to vital signs in infected patients identified consistent subphenotypes with distinct clinical outcomes and treatment responses. Time series clustering with distinct computational approaches demonstrate similar performance and significant agreement in the resulting subphenotypes.

ShVEEGc: EEG Clustering With Improved Cosine Similarity-Transformed Shapley Value

As the number of unlabeled or mislabeled electroencephalogram (EEG) increases dramatically in such applications as cerebral disease diagnosis, rehabilitation, and brain-computer interfaces, the supervised approaches that require labels or markers become inapplicable. Unfortunately, there are few reports on unsupervised studies for unlabeled EEG data, especially for unlabeled EEG clustering. To address the challenging task, we propose an effective approach named ShVEEGc for EEG clustering inspired by an improved Shapley value in cooperative game theory. The idea of ShVEEGc is first utilizing an improved cosine similarity to measure the correlations of EEG data and then calculating the improved Shapley value based on the inherent connection between unlabeled EEG data, which considers both global connections and local relationships potentially hidden in EEG data. Thus, ShVEEGc not only has good anti-interference ability but also can mine potential relationships among unlabeled EEG data. The comparison experiments with fourteen state-of-the-art EEG time series clustering algorithms on eleven real-world EEG datasets with four standard evaluation criteria demonstrate the efficacy and superiority of ShVEEGc for EEG clustering. Besides, the discussion on the impact of several different similarity measures on ShVEEGc also illustrates that the improved cosine similarity proposed in this paper is more suitable for EEG data.

Temporal dynamics clustering for analyzing cell behavior in mobile networks

In recent years, management and resource allocation of mobile networks have become crucial because of a dramatic growth of mobile traffic. For Internet service providers, resource allocation is a challenging problem due to the inhomogeneous distributed nature of user demand, which increases the difficulty of configuring base stations in complicated urban environments. To manage mobile networks in an efficient way, understanding the typical usage behavior is very valuable for network operators. This paper aims at providing tools to analyze the behavior of mobile network cells in an urban environment. To achieve this goal, we propose a new time series clustering algorithm denoted as temporal dynamics clustering. Unlike conventional clustering approaches, this algorithm generates a new representation of time series by summarizing the distribution of the sequence of differences; in this way, temporal dynamics clustering creates clusters based on the variability of time series. We apply the proposed algorithm on a real-world mobile network dataset, showing its superior performance and much faster running time with respect to conventional methods such as hierarchical clustering and K-medoids. We also train neural networks on the clusters generated by temporal dynamics clustering, and the results show that the forecasting difficulty of time series is closely related to its temporal dynamics. Using this approach, we have analyzed the cell behavior with two weeks observation collected in Turin, and obtained interpretable results, in that the behavior of each cell is strongly related to the characteristics of the area and the related human activity. We also study the cell behavior in different time slots by considering time-dependent characteristic of human activity. Through experiments, we show the effectiveness of the proposed algorithm at providing a deeper understanding of traffic usage patterns in intricate urban environments.

Time series clustering method with cluster validation to identify unknown local cell conditions in the aluminum reduction cell

It is important to identify local cell conditions in the aluminum reduction cell for cell control. However, in the actual aluminum electrolysis process, local cell conditions are unknown based on traditional measurements. Fortunately, the appearance of anode current signals as a new process variable makes it possible to identify local cell conditions. In this paper, without any prior knowledge on local cell conditions, a new time series clustering method with cluster validation is proposed to identify unknown local cell conditions. First, the nature of anode current signals is analyzed and a new process variable named the pseudo-resistance of an anode–cathode path (Rpseudo) is transformed from the anode current signals, which can provide more insights to the aluminum electrolysis process. Then, a time series clustering algorithm that combines fuzzy c-means with the dynamic time warping distance is used to cluster the anode current signals and Rpseudo. Finally, a new clustering validity index is designed and a cluster validation process is proposed to determine the optimal number of clusters. Moreover, different anode current signals for different local cell conditions such as normal working, anode deformation, anode effect and anode slippage are deliberately selected to verify the proposed method. The experimental results show that the proposed method can effectively identify local cell conditions.

DYNAMITE: Dynamic aggregation of mutually‐connected points based clustering algorithm for time series data

Time‐series clustering typically entails clustering similar patterns across various time scales or comparing various point trajectories. However, this study emphasizes to group data points based on their motions and forecast how the clusters will evolve across a number of immediate time frames. To achieve this, we propose a DYNamic Aggregation of Mutually‐connected poInts clusTEring (DYNAMITE) based clustering algorithm for time series. DYNAMITE is based on the interaction between points in a time series and it majorly consists of three components: (1) cluster initialization; (2) calculation of mutually connected points; and (3) cluster updating.

A Fast Weighted Fuzzy C-Medoids Clustering for Time Series Data Based on P-Splines.

The rapid growth of digital information has produced massive amounts of time series data on rich features and most time series data are noisy and contain some outlier samples, which leads to a decline in the clustering effect. To efficiently discover the hidden statistical information about the data, a fast weighted fuzzy C-medoids clustering algorithm based on P-splines (PS-WFCMdd) is proposed for time series datasets in this study. Specifically, the P-spline method is used to fit the functional data related to the original time series data, and the obtained smooth-fitting data is used as the input of the clustering algorithm to enhance the ability to process the data set during the clustering process. Then, we define a new weighted method to further avoid the influence of outlier sample points in the weighted fuzzy C-medoids clustering process, to improve the robustness of our algorithm. We propose using the third version of mueen’s algorithm for similarity search (MASS 3) to measure the similarity between time series quickly and accurately, to further improve the clustering efficiency. Our new algorithm is compared with several other time series clustering algorithms, and the performance of the algorithm is evaluated experimentally on different types of time series examples. The experimental results show that our new method can speed up data processing and the comprehensive performance of each clustering evaluation index are relatively good.

A new shape-based clustering algorithm for time series

Time series clustering is a research hotspot in data mining. Most of the existing clustering algorithms combine with the classical distance measure which ignore the offset of sequence shape. As a result, shape-based clustering algorithms are becoming increasingly popular. On the majority of data sets, the most representative shape-based clustering algorithm, KShape, which defines a shape-based distance with shift invariance, has been shown to outperform other algorithms.In this paper, we propose a new shape-based clustering algorithm named Fractional Order Shape-based k-cluster(FrOKShape), which defines a multi-variable shape-based distance by normalized fractional order cross-correlation and uses the DTW Barycenter Averaging (DBA) as a center computation strategy. Our distance exhibits excellent shape shift deviating properties and good compatibility integrated with a variety of existing clustering center strategies so that it can provide more potential good results. Experiments show that combining our distance with a traditional clustering algorithm produces excellent clustering indicators. In a series of comparative experiments, FrOKShape also exhibits a comparable result to the existing better shape-based clustering algorithm KShape.

An Interpretable Deep Transfer Learning-Based Remaining Useful Life Prediction Approach for Bearings With Selective Degradation Knowledge Fusion

This article tries to answer the two questions of bearings’ remaining useful life (RUL) prediction with deep transfer learning: what bearing data in the source domain contribute more to transfer learning and how to quantify such contribution? From the perspective of sample-based interpretability, this article proposes a new deep transfer learning approach of RUL prediction. First, a new time series clustering algorithm based on multiscale degradation similarity is proposed. Comprehensively considering the geometry and tendency characteristics of degradation sequences, this algorithm can divide the source domain into multiple subsource domains. Second, a significance metric, named transfer domain validity index (T-DVI), is built to quantify the contribution of each subsource domain to transfer learning in terms of degradation similarity. Furthermore, a new stacked long short-term memory model with selective transfer learning is constructed. Running with the obtained T-DVIs, this model builds two new transfer strategies of weighted initialization and adaptive freezing to improve the transfer effect of degradation knowledge for RUL prediction. Experimental results on the two bearing datasets prove the effectiveness of the proposed approach. More importantly, the proposed approach makes the transfer learning process interpretable via identifying the significance of the bearings data, which helps transfer useful degradation knowledge and improve the RUL prediction performance as well.

Dynamic driving intention recognition of vehicles with different driving styles of surrounding vehicles

Driving behaviour of surrounding vehicles have impacts on the security of ego vehicle. However, many vehicles do not use the turn signal in advance as a warning when changing lanes, which often leads to traffic accidents. Therefore, in the case of Internet of Vehicles (IoV) technology has not been popularized, recognizing the driving intention of surrounding vehicles timely and accurately through effective method is of great significance to ego vehicle. In this paper, a dynamic driving intention recognition method for surrounding vehicles based on adaptive multi-dimension continuous Gaussian mixture-HMM (AMCGM-HMM) is proposed. Considering that driving styles of surrounding vehicles are different, a feature-based multi-dimensional time series clustering algorithm is adopted to cluster the lane changing data samples and the results are used as the input of AMCGM-HMM. Parameters that may affect the recognition accuracy are optimized by multi-objective particle swarm optimization (MOPSO) algorithm to realize the adaptive of the method. The proposed method is verified on NGSIM dataset, and the application of this model is tested on the PreScan simulation platform combined with adaptive cruise control (ACC) system. Experimental results show that the proposed method can identify the driving intention of surrounding vehicles in advance and has high accuracy.

An analytical framework based on the recency, frequency, and monetary model and time series clustering techniques for dynamic segmentation

Nowadays, banks use data mining and business intelligence tools and techniques to analyze their customers’ behavior. Customer segmentation is a widely adopted analytical tool to identify distinct groups of customers. Most studies have adopted a static segmentation approach, leading to missing important trends and patterns. In this study, we propose a framework that represents each customer behavior as a time-series sequence of the Recency, Frequency, and Monetary variables, and then exploits time-series clustering algorithms to carry out customer segmentation. The framework consists of state-of-the-art clustering algorithms, including hierarchical, spectral, and k-shape algorithms to divide customers into homogeneous groups and implement point of sale devices transaction data of grocery and appliance retailers. We divide customers into four segments, analyze the behavioral trends, and provide marketing suggestions for each segment. Our results show that, in terms of the computed validity indices, the best clustering model for grocery retailers is reached using hierarchical clustering with the Complexity-Invariant Distance measure, and for appliance retailers, the best segmentation is achieved by applying the spectral clustering with the Complexity-Invariant Distance measure.

A novel short-term load forecasting framework based on time-series clustering and early classification algorithm

With the development of data-driven models, extracting information from historical data for better energy forecasting is critically important for energy planning and optimization in buildings. Feature engineering is a key factor in improving the performance of forecasting models. Adding load pattern labels for different daily energy consumption patterns resulting from different time schedules and weather conditions can help improve forecasting accuracy. Traditionally, pattern labeling focuses mainly on finding a day similar to the forecasting day based on calendar or other information, such as weather conditions. The most intuitive approach for dividing historical time-series load into patterns is clustering; however, the pattern cannot be determined before the load is known. To address this problem, this study proposes a novel short-term load forecasting framework integrating an early classification algorithm that uses a stochastic algorithm to predetermine the load pattern of a forecasting day. In addition, a hybrid multistep method combining the strengths of single-step forecasting and recursive multistep forecasting is integrated into the framework. The proposed framework was validated through a case study using actual metered data. The results demonstrate that the early classification and proposed labeling strategy produce satisfactory forecasting accuracy and significantly improve the forecasting performance of the LightGBM model.

Forecasting retail sales based on cheng fuzzy time series and particle swarm optimization clustering algorithm

Use of the conventional forecasting method, which is based on trend data with average sales in the last few months, results inaccurate forecasting due to a large difference in data, this is the same as fuzzy forecasting with the same interval length or static. Therefore, this paper recommends using the Cheng forecasting method combined with the Particle Swarm Optimization (PSO) algorithm. We use an artificial intelligence, i.e., PSO algorithm to set non-static length of intervals each cluster on Cheng method. The comparison of this method yields a better root mean square error (RMSE) value for each cluster on the recommended method.