Multivariate Demand Forecasting for Rental Bike Systems Based on an Unobserved Component Model

Massive bike-sharing systems (BSS) usage and performance data have been collected for years over various locations. Nevertheless, researchers encountered several challenges while dealing with massive BSS data. The challenges that could be enhanced in the previous studies are 1) reducing high dimensionality and noise of BSS time series data and 2) extracting informative usage patterns out of massive BSS data. This paper extracts patterns and reduce data dimensions of BSS usage by exploring time series representation and clustering of BSS usage data. A reduced dimension allows us to efficiently approximate the BSS usage with reasonable accuracy, which can be further used for bike usage clustering, classification and prediction. We employ a non-data adaptive representation technique -Discrete Wavelet Transform (DWT) to reduce dimensionality and filter out random errors of the raw time series. Time series are clustered using k-means based on similarities measured by Dynamic Time Warping (DTW) and prototypes computed using DTW barycenter averaging (DBA). The proposed approaches are applied on a 3-month bike usage dataset acquired on the BSS of Chicago. The analysis results show that DWT can effectively reduce dimensionality, filter out random errors and reveal the main characteristics of the raw time series. The clustering approach offers the ability to differentiate and discover bike usage patterns across different stations.

Bike Sharing Systems (BSS) have become a key solution for urban mobility, reducing traffic-related CO2 emissions. However, managing BSS poses challenges that require data-driven solutions, particularly for understanding their global behavior and forecasting their evolution. These dynamics arise from the interaction among users, companies, dock stations, and city policies, influenced by sociological and infrastructure-based factors. This paper proposes a novel clustering methodology to analyze BSS data across multiple cities. By clustering station-day tuples instead of aggregating statistics, our approach captures seasonal patterns, special events, and weekday/weekend differences. Using Pearson Correlation as a distance metric, it remains robust across different station sizes and system scales. Trained on three European BSS and evaluated across six cities from 4 different countries, our model uncovers meaningful patterns such as work, residential, and leisure areas, as well as seasonal changes even in systems not used in the training process. These insights enhance BSS management, expansion, and decision-making, with applications in monitoring, anomaly detection, and demand prediction.

The paper aims to quantify the daily bicycle traffic volume based on the data from bicycle automatic counters and bikeshare system (BSS) trip data taking into account the impact of weather conditions. Changes in bicycle traffic were assessed based on coefficient of variation of bicycle volume and number of BSS trips. The research was conducted in the city of Cracow (Poland) for 2 years. The volume at 13 locations with bicycle automatic counters and a daily number of trips on public bicycles were avaialble. Data on daily weather conditions were collected from a weather station located in the city.The monthly coefficients of variation of the number of BSS trips and cyclists’ volume in analyzed locations are not statistically significantly different. It means data from BSS, easily to obtain, can be used to assess variation in cyclists volume. Mean daily air temperature, daily rainfall, public and school holidays had a statistically significant impact on bicycle volume. Models have shown that BSS trip data can be successfully used to evaluate the impact of weather conditions on bicycle volume. Additionally, based on the developed models, reliable coefficient of variations for Annual Average Daily Bicycle Traffic of BSS users and all cyclists were indicated.

The imbalance between the supply and demand of shared bikes is prominent in many urban rail transit stations, which urgently requires an efficient vehicle deployment strategy. In this paper, we propose an integrated model to optimize the deployment of shared bikes around urban rail transit stations, incorporating a seasonal autoregressive integrated moving average with long short-term memory (SARIMA-LSTM) hybrid model that is used to predict the heterogeneous demand for shared bikes in space and time. The shared bike deployment strategy was formulated based on the actual deployment process and under the principle of cost minimization involving labor and transportation. The model is applied using the big data of shared bikes in Xicheng District, Beijing. Results show that the SARIMA-LSTM hybrid model has great advantages in predicting the demand for shared bikes. The proposed allocation strategy provides a new way to solve the imbalance challenge between the supply and demand of shared bikes and contributes to the development of a sustainable transportation system.

The relationship between the Beveridge–Nelson decomposition and other permanent–transitory decompositions that are popular in economics

Bicycle-sharing systems (BSSs) have become a prominent feature of the transportation network in many cities. Along with the boom of BSSs, cities face the challenge of bicycle unavailability and dock shortages. It is essential to conduct rebalancing operations, the success of which largely depend on users' demand prediction. The objective of this study is to develop users' demand prediction models based on the rental data, which will serve rebalancing operations. First, methods to collect and process the relevant data are presented. Bicycle usage patterns are then examined from both trip-based aspect and station-based aspect to provide some guidance for users' demand prediction. After that, the methodology combining cluster analysis, a back-propagation neural network (BPNN), and comparative analysis is proposed to predict users' demand. Cluster analysis is used to identify different service types of stations, the BPNN method is utilized to establish the demand prediction models for different service types of stations, and comparative analysis is employed to determine if the accuracy of the prediction models is improved by making a distinction among stations and working/nonworking days. Finally, a case study is conducted to evaluate the performance of the proposed methodology. Results indicate that making a distinction among stations and working/nonworking days when predicting users' demand can improve the accuracy of prediction models.

Forecasting time series data is an integral component for management, planning and decision making. Following the Big Data trend, large amounts of time series data are available from many heterogeneous data sources in more and more applications domains. The highly dynamic and often fluctuating character of these domains in combination with the logistic problems of collecting such data from a variety of sources, imposes new challenges to forecasting. Traditional approaches heavily rely on extensive and complete historical data to build time series models and are thus no longer applicable if time series are short or, even more important, intermittent. In addition, large numbers of time series have to be forecasted on different aggregation levels with preferably low latency, while forecast accuracy should remain high. This is almost impossible, when keeping the traditional focus on creating one forecast model for each individual time series. In this paper we tackle these challenges by presenting a novel forecasting approach called cross-sectional forecasting. This method is especially designed for Big Data sets with a multitude of time series. Our approach breaks with existing concepts by creating only one model for a whole set of time series and requiring only a fraction of the available data to provide accurate forecasts. By utilizing available data from all time series of a data set, missing values can be compensated and accurate forecasting results can be calculated quickly on arbitrary aggregation levels.

Chapter 10 - Time Series Forecasting

A novel featurization methodology using JaGen algorithm for time series forecasting with deep learning techniques

Spatial outliers are objects having a behavior sig- nificantly different from their spatial neighbors, in a context where neighbors are heavily correlated. Moran scatterplot is a well-known method that exploits similarity between neighbors in order to detect spatial outliers. In this paper, we proposed first an improved version of Moran scatterplot, using a robust distance metric called Gower's similarity. We used the new version of Moran scatterplot to study the homogeneity of the Parisian bike sharing system (Velib). We carried out different experiments on a real dataset issued from the Velib system. We identified many spatial outliers stations, very different from their neighboring stations (often with much more available bikes or with much more empty docks during the day). Then, we designed and tested a new method that globally improves the distribution of the resources (bikes and docks) among bike stations. This method is motivated by the existence of spatial outliers stations. It relies on a local small change in users behaviors, by adapting their trips to resources' availability around their departure and arrival stations. Results show that, even with a partial users collaboration, the proposed method enhances significantly the global homogeneity of the bike sharing system and therefore the users' satisfaction.

Hybrid time-aligned and context attention for time series prediction

It is important yet challenging to perform accurate and interpretable time series forecasting. Though deep learning methods can boost forecasting accuracy, they often sacrifice interpretability. In this paper, we present a new scheme of series saliency to boost both accuracy and interpretability. By extracting series images from sliding windows of the time series, we design series saliency as a mixup strategy with a learnable mask between the series images and their perturbed versions. Series saliency is model agnostic and performs as an adaptive data augmentation method for training deep models. Moreover, by slightly changing the objective, we optimize series saliency to find a mask for interpretable forecasting in both feature and time dimensions. Experimental results on several real datasets demonstrate that series saliency is effective to produce accurate time-series forecasting results as well as generate temporal interpretations.

Oilseeds have been the backbone of India’s agricultural economy since long. Oilseed crops play the second most important role in Indian agricultural economy, next to food grains, in terms of area and production. Oilseeds production in India has increased with time, however, the increasing demand for edible oils necessitated the imports in large quantities, leading to a substantial drain of foreign exchange. The need for addressing this deficit motivated a systematic study of the oilseeds economy to formulate appropriate strategies to bridge the demand-supply gap. In this study, an effort is made to forecast oilseeds production by using Autoregressive Integrated Moving Average (ARIMA) model, which is the most widely used model for forecasting time series. One of the main drawbacks of this model is the presumption of linearity. The Group Method of Data Handling (GMDH) model has also been applied for forecasting the oilseeds production because it contains nonlinear patterns. Both ARIMA and GMDH are mathematical models well-known for time series forecasting. The results obtained by the GMDH are compared with the results of ARIMA model. The comparison of modeling results shows that the GMDH model perform better than the ARIMA model in terms of mean absolute error (MAE), mean absolute percentage error (MAPE), and root mean square error (RMSE). The experimental results of both models indicate that the GMDH model is a powerful tool to handle the time series data and it provides a promising technique in time series forecasting methods.

The volatility of the housing price based on demand and supply in the housing market reflects the structural features of the housing market. Consequently, decomposing the housing market into trends and cycles through econometric models will make it possible to deduce the nature of the housing market in the long term, in addition to provide hindsight for real estate portfolio asset allocation in a more systematic way.BRBR The paper addresses the housing price behavior through univariate time series analysis. The Unobserved-Component (UC) model and the Beveridge-Nelson (BN) decomposition model are applied to analyze the factors for the volatility of housing price with the trends and cycles. Specifically, the UC model is analyzed by UC-Zero(0) model assuming the covariance between the trends and the cycles, which are unobserved variables, is zero (0), and the UC-UR (UC-UnRestricted) model, which is free from the restraint of the covariance between the trends and cycles. The estimation is based on the Kalman Filter. The study analyzes how the volatility in the domestic housing price index was influenced by the trends and cycles, which are inherent variables, by comparing the estimated values of the UC-0 and UC-UR model with the BN decomposition model. Domestic housing prices have plunged two times: the 1997 Asian financial crisis and the 2008 Global financial crisis. However, the analysis data in this research include only the 2008 Global financial crisis.BRBR The results are summarized as follows: First, the volatility of the domestic housing price index can be explained by the UC-0 model. As explained by the UC-0 model with the assumption of the covariance between trend and cycles zero(0), the domestic housing price index has not shown any correlation between the trend and cycle, and thus it implies that the trends mainly explain the volatility in the housing price index. Second, as the trends mainly explain the volatility in housing prices, domestic housing prices do not follow a random walk under the efficient market hypothesis. Hence, the price volatility can be seen as predictable, and the right-upward price trend that has been lasted for a considerable period of time can be seen as a result of reflecting adaptive expectations of market participants. Lastly, institutional investors such as pension funds are increasing their share of real estate and alternative investments, and the share of real estate is increasing in the asset market. The fact that inefficient housing market, and the housing price return as seen through the housing price analysis follow a very differential distribution from the normal distribution would mean that as a factor hindering investors

В работе предложены математические модели прогнозирования нестационарных временных рядов с учётом внешних факторов и методы их структурной идентификации, основанные на совместном использовании многомерного варианта метода «Гусеница»-SSA и моделей SARIMAX, расширенных для прогнозирования временных рядов с несколькими сезонными составляющими и с учётом нескольких экзогенных переменных. Экспериментальные результаты показывают высокую эффективность предложенных моделей прогнозирования при выборе подходящих структурных параметров в сравнении с моделями SARIMAX.