Time Series Data Research Articles

Unravelling the Effects of Climate Change on Agriculture of Pakistan: An Exploratory Analysis

The agricultural sector of Pakistan, the backbone of the economy, has experienced extreme changes in climate over the past years. Varying temperatures, rainfall patterns, and carbon emissions have drastically affected the country's agricultural productivity. This study focuses on the exploratory analysis of various factors of climate change in the agricultural sector of Pakistan and involves a descriptive analysis of three variables: temperature, rainfall, and carbon emissions. Time series data on monthly temperature and rainfall was taken from the World Bank from the year 1990 to 2018. At the same time, carbon emissions were studied on a yearly basis. The analysis showed that the country had experienced an increase in temperature and carbon emissions during the period of analysis. However, the decreasing trends of rainfall during most of the months explored the deteriorating agricultural production and food insecurity in the country. Therefore, the government should employ innovative techniques to harvest rainwater for agricultural development, aiming to reduce carbon emissions.

External Debt and Economic Growth in Sierra Leone: A Threshold Analysis

The main focus of this study is to investigate the impact of external debt on economic growth and determine the threshold level of external debt that is favourable for economic growth in Sierra Leone, within the framework of a non-linear model (quadratic function). The study employs the Ordinary Least Squares estimation technique with annual time series data for the period 1980 to 2022. The empirical findings reveal that, lagged GDP, external debt, external debt squared and investment are the main drivers of economic growth in Sierra Leone. Specifically, the result reveals that external debt, investment and lagged GDP have a positive impact on growth whilst, external debt squared has a harmful impact on growth. Furthermore, the study establishes a nonlinear relationship between external debt and economic growth. The result indicates that the optimal threshold level of external debt (as a % of GDP) in Sierra Leone is 115%. The findings reveal that, external debt below the threshold level is growth enhancing. However, the accumulation of external debt beyond the threshold level of 115% is detrimental to growth. The diagnostic result confirms that about 76% of the variation in economic growth is explained by the regressors (independent variables). Also, the granger causality test result suggests a unidirectional causality from external debt to economic growth with no reverse causality, indicating that external debt granger caused economic growth at the 5% level. Based on the findings, the study recommends that, government should develop a robust debt management policy, to keep debt levels within sustainable limits and also cultivate an economic culture of transparency, on the issue of debt management and contract negotiations.

Exploring the training results of machine learning models using different batch sizes and epochs: A case study with GNSS time series data

Exploring the training results of machine learning models using different batch sizes and epochs: A case study with GNSS time series data

Thinking in Categories: A Survey on Assessing the Quality for Time Series Synthesis

Time series data are widely used and provide a wealth of information for countless applications. However, some applications are faced with a limited amount of data, or the data cannot be used due to confidentiality concerns. To overcome these obstacles, time series can be generated synthetically. For example, electrocardiograms can be synthesized to make them available for building models to predict conditions such as cardiac arrhythmia without leaking patient information. Although many different approaches to time series synthesis have been proposed, evaluating the quality of synthetic time series data poses unique challenges and remains an open problem, as there is a lack of a clear definition of what constitutes a “good” synthesis. To this end, we present a comprehensive literature survey to identify different aspects of synthesis quality and their relationships. Based on this, we propose a definition of synthesis quality and a systematic evaluation procedure for assessing it. With this work, we aim to provide a common language and criteria for evaluating synthetic time series data. Our goal is to promote more rigorous and reproducible research in time series synthesis by enabling researchers and practitioners to generate high-quality synthetic time series data.

Regressor cascading for time series forecasting

Time series forecasting is the process of predicting future values of a time series based on its historical data patterns. It is a critical task in many domains, including finance, supply chain management, the environment, and more as accurate forecasts can help businesses and organizations make better decisions and improve their metrics. Although there have been significant advances in time series forecasting systems, thanks to the development of new machine learning algorithms, hardware improvements, and the increasing availability of data, it remains a challenging task. Common pitfalls, especially of single-model approaches include susceptibility to noise and outliers and inability to handle non-stationary data, which can lead to inaccurate and non-robust forecasts. Model-combining approaches, such as averaging the results of multiple predictors to produce a final forecast, are commonly used to mitigate such issues. This work introduces a novel application of Cascade Generalization or Cascading for time series forecasting, where multiple predictors are used sequentially, with each predictor’s output serving as additional input for the next. This methodology aims to overcome the limitations of single-model forecasts and traditional ensembles by incorporating a progressive learning mechanism. We adapt Cascade Generalization specifically for time series data, detailing its implementation and potential for handling complex, dynamic datasets. Our approach was systematically evaluated against traditional two-model averaging ensembles across ten diverse datasets, employing the Root Mean Square Error (RMSE) metric for performance assessment. The results revealed that cascading tends to outperform voting ensembles in most cases. This consistent trend suggests that cascading can be considered a reliable alternative to voting ensembles, showcasing its potential as an effective strategy for improving time series forecasting across a wide range of scenarios.

Data assimilation with machine learning for constructing gridded rainfall time series data to assess long-term rainfall changes in the northeastern regions in India

ABSTRACT Data scarcity and unavailability of observed rainfalls in the northeastern states of India limit prediction of extreme hydro-climatological changes. To fill this gap, a data assimilation approach has been applied to re-construct accurate high-resolution gridded (5 km2) daily rainfall data (2001–2020), which include seasonality assessment, statistical evaluation, and bias correction. Random forest (RF) and support vector regression were used to predict rainfall time series, and a comparison between machine learning and data assimilation-based gridded rainfall data was performed. Five gridded rainfall datasets, namely, Indian Monsoon Data Assimilation and Analysis (IMDAA) (12 km2), APHRODITE (25 km2), India Meteorological Department (25 km2), PRINCETON (25 km2), and CHIRPS (25 and 5 km2), have been utilized. For re-constructed rainfall datasets (5 km2), the comparative seasonality and change assessment have been performed with respect to other rainfall datasets. CHIRPS and APHRODITE datasets have shown better similarities with IMDAA. The RF and assimilated rainfall (AR) have superiority based on bias and extremity, and AR data were recognized as the best accurate data (>0.8). Precipitation change analysis (2021–2100) performed utilizing the bias corrected and downscaled CMIP6 datasets showed that the dry spells will be enhanced. Considering the CMIP6 moderate emission scenario, i.e., SSP245, the wet spell will be enhanced in future; however, when considering SSP585 (representing the extreme worst case), the wet spells will be decreased.

Spatiotemporal changes in fine particulate matter and ozone in the oasis city of Korla, northeastern Tarim Basin of China

Air pollution is a serious environmental health concern for humans and other living organisms. This study analyzes the spatial and temporal characteristics of air pollutant concentrations, changes in the degree of pollution, and the wavelet coherence of the air quality index (AQI) with pollutants in various monitoring stations. The analysis is based on long-term time series data (January 2016 to December 2023) of air pollutants (PM2.5, PM10, and O3) from Korla, an oasis city in the northeastern part of the Tarim Basin, China. The concentrations of PM2.5, PM10, and O3 in Korla showed a cyclical trend from 2016 to 2023; PM10 concentrations exhibited all-season exceedance and PM2.5 exhibited exceedance only in spring. PM2.5 and PM10 showed a seasonal distribution of spring > winter > fall > summer; O3 concentrations showed a seasonal distribution of summer > spring > fall > winter. Strong positive wavelet coherence between PM and Air Quality Index (AQI) data series suggests that the AQI data series can effectively characterize fluctuating trends in PM concentrations. Moreover, PM10 levels IV and VI were maintained at approximately 10%, indicating that sand and dust have a substantial influence on air quality and pose potential threats to the health of urban inhabitants. Based on the results of this study, future efforts must strengthen relative countermeasures for sand prevention and control, select urban greening species with anti-pollution capabilities, rationally expand urban green spaces, and restrict regulations for reducing particulate matter emissions within city areas.

Variance-Reduced Stochastic Optimization for Efficient Inference of Hidden Markov Models

Hidden Markov models (HMMs) are popular models to identify a finite number of latent states from sequential data. However, fitting them to large datasets can be computationally demanding because most likelihood maximization techniques require iterating through the entire underlying dataset for every parameter update. We propose a novel optimization algorithm that updates the parameters of an HMM without iterating through the entire dataset. Namely, we combine a partial E step with variance-reduced stochastic optimization within the M step. We prove the algorithm converges under certain regularity conditions. We test our algorithm empirically using a simulation study as well as a case study of kinematic data collected using suction-cup attached biologgers from eight northern resident killer whales (Orcinus orca) off the western coast of Canada. In both, our algorithm converges in fewer epochs, with less computation time, and to regions of higher likelihood compared to standard numerical optimization techniques. Our algorithm allows practitioners to fit complicated HMMs to large time-series datasets more efficiently than existing baselines. Supplemental materials are available online.

Higher-Order Convolutional Neural Networks for Essential Climate Variables Forecasting

Earth observation imaging technologies, particularly multispectral sensors, produce extensive high-dimensional data over time, thus offering a wealth of information on global dynamics. These data encapsulate crucial information in essential climate variables, such as varying levels of soil moisture and temperature. However, current cutting-edge machine learning models, including deep learning ones, often overlook the treasure trove of multidimensional data, thus analyzing each variable in isolation and losing critical interconnected information. In our study, we enhance conventional convolutional neural network models, specifically those based on the embedded temporal convolutional network framework, thus transforming them into models that inherently understand and interpret multidimensional correlations and dependencies. This transformation involves recasting the existing problem as a generalized case of N-dimensional observation analysis, which is followed by deriving essential forward and backward pass equations through tensor decompositions and compounded convolutions. Consequently, we adapt integral components of established embedded temporal convolutional network models, like encoder and decoder networks, thus enabling them to process 4D spatial time series data that encompass all essential climate variables concurrently. Through the rigorous exploration of diverse model architectures and an extensive evaluation of their forecasting prowess against top-tier methods, we utilize two new, long-term essential climate variables datasets with monthly intervals extending over four decades. Our empirical scrutiny, particularly focusing on soil temperature data, unveils that the innovative high-dimensional embedded temporal convolutional network model-centric approaches markedly excel in forecasting, thus surpassing their low-dimensional counterparts, even under the most challenging conditions characterized by a notable paucity of training data.

Detection of Approaching Critical Transitions in Natural Systems Driven by Red Noise

Detection of critical slowing down (CSD) is the dominant avenue for anticipating critical transitions from noisy time-series data. Most commonly, changes in variance and lag-1 autocorrelation [AC(1)] are used as CSD indicators. However, these indicators will only produce reliable results if the noise driving the system is white and stationary. In the more realistic case of time-correlated red noise, increasing (decreasing) the correlation of the noise will lead to spurious (masked) alarms for both variance and AC(1). Here, we propose two new methods that can discriminate true CSD from possible changes in the driving noise characteristics. We focus on estimating changes in the linear restoring rate based on Langevin-type dynamics driven by either white or red noise. We assess the capacity of our new estimators to anticipate critical transitions and show that they perform significantly better than other existing methods both for continuous-time and discrete-time models. In addition to conceptual models, we apply our methods to climate model simulations of the termination of the African Humid Period. The estimations rule out spurious signals stemming from nonstationary noise characteristics and reveal a destabilization of the African climate system as the dynamical mechanism underlying this archetype of abrupt climate change in the past. Published by the American Physical Society 2024

Impact of RMB internationalization on China's competitiveness in financial services trade based on the VAR model: Evidence from China-US.

An increase in a currency internationalization levels can positively impact its credibility in international economic activities, and expand the effective demand and optimize the supply structure for the country's financial service trade. In this way, a state can improve its financial service trade competitiveness in the international market. This study builds a vector autoregressive model based on time-series data of China-US financial services trade from 2010 to 2021, analyzes the impact of different quantitative indicators of RMB internationalization on this trade from the impulse response results, and validates the conclusions using various inspection methods. The results show that the increase in RMB internationalization helps to narrow the China-US financial services trade balance, but with a significant lag. And this effect is heterogeneous in different dimensions, demonstrated by the fact that the development of overseas RMB securities business is more important for the level of RMB internationalization to narrow the China-US financial services trade balance. Finally, among the specific measures to improve its financial services trade, China should focus on developing the international competitiveness of the traditional RMB deposit and loan financial sector, while the competition in the overseas market for high value-added financial businesses must also not be neglected. Furthermore, China needs to implement more targeted RMB internationalization development policies at different levels in the future to provide high-quality financial services to the rest of the world and aid in the economic recovery of the world in the "post-pandemic" era.

A novel transformer-based neural network under model interpretability for improving wheat yield estimation using remotely sensed multi-variables

Timely and accurate yield prediction before wheat harvest is of great significance for food policy formulation and national economic development. Deep learning method gains importance on crop yield estimation and growth monitoring with the rapid development of deep learning models combined with remotely sensed data. However, the problems of nonlinear parameter optimization and long-term dependence in time series data have restricted the improvement of yield estimation accuracy. Transformer model completely abandons the traditional deep learning architecture and is gradually applied to time series tasks due to the advantage in modeling long-term dependence. This study introduced a novel transformer-based deep learning framework to estimate winter wheat yield in the Guanzhong Plain utilizing remotely sensed multi-variables, namely leaf area index (LAI), fraction of photosynthetically active radiation (FPAR) and vegetation temperature condition index (VTCI). The proposed model, called SSA-LSTM-transformer (SLTF), was developed for dealing with above problems under the automatic optimization capability of the sparrow search algorithm (SSA) and the long-term memory advantage of the long short-term memory (LSTM) structure. The findings demonstrated that the SLTF model achieved better estimation accuracy at the county level (R2 = 0.72, RMSE = 488.68 kg/ha) compared with the single transformer model (R2 = 0.61, RMSE = 573.99 kg/ha). The analysis on SLTF model’s performance at sampling sites over different disasters suggested that the model can effectively learn the effects of diseases, lodge and pests on crop yield estimation, which had good generalization ability at all sampling sites (R2 = 0.45, RMSE = 738.63 kg/ha). The Shapley Additive exPlanations (SHAP) approach was employed to evaluate the relative importance of remotely sensed multi-variables to yield and to understand how each input variable influences the estimated yield for global interpretability and local interpretability of the SLTF model. It was found that FPAR played the largest roles in yield estimation with the highest importance value, and FPAR and LAI from late April to late May and VTCI from late March to mid-April were regarded as important features for the estimated yield with high importance value. In conclusion, our findings indicated the potential of SLTF model for winter wheat yield estimation, which contributes to promoting further application of remotely sensed technology for agricultural production.

MCNet: Multivariate long-term time series forecasting with local and global context modeling

Time series data typically exhibit various intra-sequence and inter-sequence correlations, resulting in intricate, intertwined dependencies, which pose challenges for accurately predicting future long-term trends. Previous studies have not fully considered the two correlations, and they also still face challenges of excessive time and memory complexity when dealing with long-term predictions. To address these challenges and establish high-precision prediction models, we propose MCNet that consists of a local branch and a global branch. The local branch aims at capturing short-term variations of intra-sequences, as well as capturing inter-sequence correlations. The global branch models long-term dependencies within sequences. Specifically, the local branch consists only of MLP module, which effectively captures short-term variations of intra-sequences by independently modeling the temporal information within and between patches of the most recent time series. Subsequently, inter-sequence dependencies are captured through the channel interaction module, which further explores more key information to improve the performance of MCNet. Meanwhile, global branch models long-term dependencies within the time series through structured global convolution. Experimental results on multiple popular long-term time series forecasting benchmarks demonstrate that MCNet outperforms state-of-the-art methods, yielding a relative improvement of 12% for multivariate time series while also being more efficient.

Temporal Attention for Few-Shot Concept Drift Detection in Streaming Data

Concept drift describes unforeseeable changes in the underlying distribution of streaming data over time. Concept drift is a phenomenon in which the statistical properties of a target domain change over time in an arbitrary way. These changes might be caused by changes in hidden variables that cannot be measured directly. With the onset of the big data era, domains such as social networks, meteorology, and finance are generating copious amounts of streaming data. Embedded within these data, the issue of concept drift can affect the attributes of streaming data in various ways, leading to a decline in the accuracy and performance of models. There is a pressing need for new models to re-adapt to the changes in streaming data. Traditional concept drift detection algorithms struggle to effectively capture and utilize the key feature points of concept drift within complex time series, thereby failing to maintain the accuracy and efficiency of the models. In light of these challenges, this study introduces a novel concept drift detection method that incorporates a temporal attention mechanism within a prototypical network. By integrating a temporal attention mechanism during the feature extraction process, our approach enhances the capability to process complex time series data, preserves temporal locality, strengthens the learning of key features, and reduces the amount of labeled data required. This method significantly improves the detection accuracy and efficiency of small sample streaming data by better capturing the local features of the data. Experiments conducted across multiple datasets demonstrate that this method exhibits comprehensive leading performance in terms of accuracy and F1-score, with excellent recall and precision, thereby validating its effectiveness in enhancing concept drift detection in streaming data.

Forecasting long-term sequences based on a seasonal and periodic-trend feature disentangled network

In this paper, we take into account different repetitive patterns of the seasonal, trendy, and complex periodic characteristics of time series data, and propose a long-term sequence forecasting model, which disentangles seasonal and periodic-trend feature representation. The proposed model consists of two stages. First stage concludes a decomposition component, which divides the temporal data into seasonal and periodic-trend terms. Second stage is a dual-branch prediction model. In one branch, we employ the Squeeze-and-Excitation (SE) attention mechanism to a 2D visual backbone block, which forms a two-level cascaded structure, to predict the multi-period attributes of seasonal term, and in the other, we use autoregression to forecast the periodic-trend term. Experiments on four benchmark datasets have shown that this proposed disentangled prediction network has more advanced prediction performance compared to the current advanced method TimesNet, resulting in an average MSE reduction of 8.1% for long-term series prediction.

Capsule Network Based on Double-layer Attention Mechanism and Multi-scale Feature Extraction for Remaining Life Prediction

The era of big data provides a platform for high-precision RUL prediction, but the existing RUL prediction methods, which effectively extract key degradation information, remain a challenge. Existing methods ignore the influence of sensor and degradation moment variability, and instead assign weights to them equally, which affects the final prediction accuracy. In addition, convolutional networks lose key information due to downsampling operations and also suffer from the drawback of insufficient feature extraction capability. To address these issues, the two-layer attention mechanism and the Inception module are embedded in the capsule structure (mai-capsule model) for lifetime prediction. The first layer of the channel attention mechanism (CAM) evaluates the influence of various sensor information on the forecast; the second layer adds a time-step attention (TSAM) mechanism to the LSTM network to weigh the contribution of different moments of the engine's whole life cycle to the prediction, while weakening the influence of environmental noise on the prediction. The Inception module is introduced to perform multi-scale feature extraction on the weighted data to capture the degradation information to the maximum extent. Lastly, we are inspired to employ the capsule network to capture important position information of high and low-dimensional features, given its capacity to facilitate a more effective rendition of the overall features of the time-series data. The efficacy of the suggested model is assessed against other approaches and verified using the publicly accessible C-MPASS dataset. The end findings demonstrate the excellent prediction precision of the suggested approach.

Comparative analysis of ARIMA and LSTM for predicting fluctuating time series data

The investigation of time series data forecasting is a critical topic within the realms of economics and business. The autoregressive integrated moving average (ARIMA) model has been prevalently utilized, notwithstanding its limitations, which include the necessity for a substantial quantity of data points and the presumption of data linearity. However, with recent developments, the long short-term memory (LSTM) network has emerged as a promising alternative, potentially overcoming these limitations. The objective of this study is to determine an effective approach for managing time series data characterized by volatility and missing values. Evaluation was conducted using RMSE for accuracy assessment, and the execution time measured using the Python Timeit library. The findings indicates that in a dataset comprising 60 data points, the LSTM model (RMSE 0.037618) surpasses the ARIMA model (RMSE 0.062667) in terms of accuracy. However, this trend reverses in a larger dataset of 228 data points, where the ARIMA model demonstrates superior accuracy (RMSE 0.006949) compared to the LSTM model (RMSE 0.036025). In scenarios with missing data, the LSTM model consistently outperforms the ARIMA model, although the accuracy of both models diminishes with an increase in the number of missing values. The ARIMA model significantly outpaces the LSTM model.

Estimation of rice yield using multivariate analysis techniques based on meteorological parameters

This study aims to develop predictive models for rice yield by applying multivariate techniques. It utilizes stepwise multiple regression, discriminant function analysis and logistic regression techniques to forecast crop yield in specific districts of Haryana. The time series data on rice crop have been divided into two and three classes based on crop yield. The yearly time series data of rice yield from 1980–81 to 2020–21 have been taken from various issues of Statistical s of Haryana. The study also utilized fortnightly meteorological data sourced from the Agrometeorology Department of CCS HAU, India. For comparing various predictive models' performance, evaluation of measures like Root Mean Square Error, Predicted Error Sum of Squares, Mean Absolute Deviation and Mean Absolute Percentage Error have been used. Results of the study indicated that discriminant function analysis emerged as the most effective to predict the rice yield accurately as compared to logistic regression. Importantly, the research highlighted that the optimum time for forecasting the rice yield is 1 month prior to the crops harvesting, offering valuable insight for agricultural planning and decision-making. This approach demonstrates the fusion of weather data and advanced statistical techniques, showcasing the potential for more precise and informed agricultural practices.

Improving dynamic prognostication of patients with hepatocellular carcinoma by node fusion algorithm equipped survival path mapping.

e13594 Background: Patients with hepatocellular carcinoma (HCC) always require routine surveillance and repeated treatment. Our previous work (1) developed a machine learning methodology called “Survival Path Mapping”, which converted the time series data into a cascading survival map to facilitate dynamic prognosis prediction and treatment planning. However, in previous work, the exponential growth of paths raised challenges of exhausting limited cohort and over-fitting risk. Novel algorithm enhancing data utilization in the Survival Path Mapping was in urgent need. Methods: A node fusion algorithm was designed. First, feature selection algorithm bifurcated each parent-nodes into temporary child-nodes, which constitutes temporary child-nodes set. Then, clustering was performed within temporary child-nodes set according to key features of each child-node. Finally, in each cluster, multiple child-nodes would be merged into one if there was no significant difference in survival. For nodes with bifurcation, a treatment recommendation algorithm was developed. A total of 19446 patients with newly diagnosed primary HCC at Sun Yat-sen University Cancer Center (SYSUCC) between 2008 and 2023 were retrospectively included as internal cohort (divided into internal train and holdout validation cohort); meanwhile 1135 patients from five hospitals were included as multicenter external validation cohort. Dedicated natural language processing program was developed to extract structured features from medical imaging reports. The time series data of patients were converted into data of time slices, with an interval of 3 months. Results: The original survival path mapping (raw-SP) consists of 463 different nodes, while the node fusion survival path mapping (fusion-SP) consisted of 84 nodes (81.8% decrease from raw-SP). The 3-year c-index of fusion-SP were 0.745 (95% CI 0.717-0.772), significantly better (p<0.001) than that of raw-SP: 0.725 (95% CI 0.696-0.755), in the internal holdout validation. Both raw-SP and fusion-SP significantly (p < 0.001) outperformed conventional BCLC and CNLC staging. In the multicenter external validation, the fusion-SP also demonstrated better performance than raw-SP and BCLC staging system. The recommended therapy calculated based fusion-SP system for patients receiving second-line or subsequent treatment had superior or non-inferior efficacy compared to BCLC guideline’s recommendation. Conclusions: The proposed node fusion algorithm solved exponential growth of paths in previous survival path algorithm, improving the accuracy and external validity of dynamic prognosis prediction for HCC patients. The fusion-SP model also retained its transparency and could be used to guide dynamic treatment planning, with its effectiveness in treatment guidance evaluated in prospective trials in the future. 1. Nat Commun 9, 2230; 2018.

ContrAttNet: Contribution and attention approach to multivariate time-series data imputation

ABSTRACT The imputation of missing values in multivariate time-series data is a basic and popular data processing technology. Recently, some studies have exploited Recurrent Neural Networks (RNNs) and Generative Adversarial Networks (GANs) to impute/fill the missing values in multivariate time-series data. However, when faced with datasets with high missing rates, the imputation error of these methods increases dramatically. To this end, we propose a neural network model based on dynamic contribution and attention, denoted as ContrAttNet. ContrAttNet consists of three novel modules: feature attention module, iLSTM (imputation Long Short-Term Memory) module, and 1D-CNN (1-Dimensional Convolutional Neural Network) module. ContrAttNet exploits temporal information and spatial feature information to predict missing values, where iLSTM attenuates the memory of LSTM according to the characteristics of the missing values, to learn the contributions of different features. Moreover, the feature attention module introduces an attention mechanism based on contributions, to calculate supervised weights. Furthermore, under the influence of these supervised weights, 1D-CNN processes the time-series data by treating them as spatial features. Experimental results show that ContrAttNet outperforms other state-of-the-art models in the missing value imputation of multivariate time-series data, with average 6% MAPE and 9% MAE on the benchmark datasets.