The ecoinvent database version 3 (part I): overview and methodology

Purpose

Good background data are an important requirement in LCA. Practitioners generally make use of LCI databases for such data, and the ecoinvent database is the largest transparent unit-process LCI database worldwide. Since its first release in 2003, it has been continuously updated, and version 3 was published in 2013. The release of version 3 introduced several significant methodological and technological improvements, besides a large number of new and updated datasets. The aim was to expand the content of the database, set the foundation for a truly global database, support regionalized LCIA, offer multiple system models, allow for easier integration of data from different regions, and reduce maintenance efforts. This article describes the methodological developments.

Methods

Modeling choices and raw data were separated in version 3, which enables the application of different sets of modeling choices, or system models, to the same raw data with little effort. This includes one system model for Consequential LCA. Flow properties were added to all exchanges in the database, giving more information on the inventory and allowing a fast calculation of mass and other balances. With version 3.1, the database is generally water-balanced, and water use and consumption can be determined. Consumption mixes called market datasets were consistently added to the database, and global background data was added, often as an extrapolation from regional data.

Results and discussion

In combination with hundreds of new unit processes from regions outside Europe, these changes lead to an improved modeling of global supply chains, and a more realistic distribution of impacts in regionalized LCIA. The new mixes also facilitate further regionalization due to the availability of background data for all regions.

Conclusions

With version 3, the ecoinvent database substantially expands the goals and scopes of LCA studies it can support. The new system models allow new, different studies to be performed. Global supply chains and market datasets significantly increase the relevance of the database outside of Europe, and regionalized LCA is supported by the data. Datasets are more transparent, include more information, and support, e.g., water balances. The developments also support easier collaboration with other database initiatives, as demonstrated by a first successful collaboration with a data project in Québec. Version 3 has set the foundation for expanding ecoinvent from a mostly regional into a truly global database and offers many new insights beyond the thousands of new and updated datasets it also introduced.

     

Electric passenger car transport and passenger car life cycle inventories in ecoinvent version 3

Purpose

Due to the large environmental challenges posed by the transport sector, reliable and state-of-the art data for its life cycle assessment is essential for enabling a successful transition towards more sustainable systems. In this paper, the new electric passenger car transport and vehicle datasets, which have been developed for ecoinvent version 3, are presented.

Methods

The new datasets have been developed with a strong modular approach, defining a hierarchy of datasets corresponding to various technical components in the vehicle. A vehicle is therefore modelled by linking together the various component datasets. Also, parameters and mathematical formulas have been introduced in order to define the amount of exchanges in the datasets through common transport and vehicle characteristics. This supports users in the choice of the amount of exchanges and enhances the transparency of the dataset.

Results

The new transport dataset describes the transport over 1 km with a battery electric passenger car taking into account the vehicle production and end of life, the energy consumption due to the use phase, non-exhaust emissions, maintenance and road infrastructure. The dataset has been developed and is suitable for a compact class vehicle.

Conclusions

A new electric passenger car transport dataset has been developed for version 3 of the ecoinvent database which exploits modularisation and parameters with the aim of facilitating users in adapting the data to their specific needs. Apart from the direct use of the transport dataset for background data, the various datasets for the different components can also be used as building blocks for virtual vehicles.

     

Life cycle inventories of electricity generation and power supply in version 3 of the ecoinvent database—part II: electricity markets

Purpose

Representative, consistent and up-to-date life cycle inventories (LCI) of electricity supply are key elements of ecoinvent as an LCI background database since these are often among the determining factors with regard to life cycle assessment (LCA) results. ecoinvent version 3 (ev3) offers new LCI data of power supply (electricity markets) in 71 geographies. This article gives an overview of these electricity markets and discusses new ecoinvent features in the context of power supply.

Methods

The annual geography- and technology-specific electricity production for the year 2008 specifies the technology shares on the high-, medium- and low-voltage level electricity markets. Data are based on IEA statistics. Different voltage levels are linked by transformation activities. Region-specific electricity losses due to power transmission and voltage transformation are considered in the market and transformation activities. The majority of the 71 power markets are defined by national boundaries. The attributional ecoinvent system model in ev3 with linking to average current suppliers results in electricity markets supplied by all geography-specific power generation technologies and electricity imports, while the consequential system model generates markets only linked to unconstrained suppliers.

Results and discussion

The availability of LCI data for 71 electricity markets in ev3 covering 50 countries reduces the “Rest-of-the-World” electricity supply not covered by country- or region-specific inventories to 17 % for the year 2008. Specific power supply activities for all countries contributing more than 1 % to global electricity production are available. The electricity markets show large variations concerning contributions from specific technologies and energy carriers. Imports can substantially change the national/regional power mix, especially in small markets. Large differences can also be observed between the electricity markets in the attributional and the consequential database calculation. Region-specific total power losses between production on the high voltage level and consumer on the low voltage level are on the order of 2.5–23 %.

Conclusions

Electricity supply mixes (electricity markets) in the ecoinvent database have been updated and substantially extended for v3. Inventories for electricity supply in all globally important economies are available with geography-specific technology and market datasets which will contribute to increasing quality and reducing uncertainties in LCA studies worldwide and to allow more accurate estimation of environmental burdens from global production chains. Future work should focus on improving the details of country-specific data, implementation of more countries into the database, splitting of large countries into smaller regions and on developing a more sophisticated approach specifying country-specific electricity mixes in consequential system models.

     

A trade-based method for modelling supply markets in consequential LCA exemplified with Portland cement and bananas

Purpose

This study proposes a method based on the analysis of trade networks over time for modelling the marginal supply of products in consequential life cycle assessment (LCA). It aims at increasing the geographical granularity of markets, accuracy of transport distances and modes and material losses during transit by creating country-specific markets, instead of region-based supply-origin markets as currently proposed by ecoinvent. It leads to a better consideration of the environmental weight of trade following a change in demand on a local market and may serve as an inspirational basis for future releases of consequential life cycle inventory (LCI) databases.

Methods

The method uses ecoinvent v.3.3 as a support LCI database and two distinct traded products: bananas and grey Portland cement. Each country involved in the trade of a said product has a corresponding market created in the LCI database. The behavior of market to a marginal change in internal demand is modelled after its marginal trading preferences: it can either affect local production, imports, exports or a mix of the first two. Markets are linked to one another based on the linear regression analysis of their historical trade relations. The inventories that follow an increase in demand of 1000 kg of bananas and grey Portland cement are calculated for each market involved in their trade and are environmentally characterized and compared to the generic region-based market datasets provided by ecoinvent to assess the gains in accuracy through a higher geographical granularity. Furthermore, the characterized inventories of the markets for bananas are compared to a parallel scenario where transport distances are kept to a minimum using the shortest path method. It isolates the environmental burden associated to the utility maximization of the demand.

Results and discussion

When comparing the characterized impacts of country-specific markets with the generic ecoinvent market datasets, disparities in results appear. They highlight the importance of transport induced by demand displacement and losses of material during transport, both being the consequences of the extent a given market decides to be supplied directly from producing markets at the margin. These are aspects that may go unaccounted for when using generic regional markets. Second, optimizing transport distances for each market decreases the environmental impacts for most categories by more than 70%.

Conclusions

This study shows there is a need for modelling and understanding market relations to more accurately define the role of trade, supply chain efficiency and import policies in LCA.

     

Life cycle inventories of electricity generation and power supply in version 3 of the ecoinvent database—part I: electricity generation

Purpose

Life cycle inventories (LCI) of electricity generation and supply are among the main determining factors regarding life cycle assessment (LCA) results. Therefore, consistency and representativeness of these data are crucial. The electricity sector has been updated and substantially extended for ecoinvent version 3 (v3). This article provides an overview of the electricity production datasets and insights into key aspects of these v3 inventories, highlights changes and describes new features.

Methods

Methods involved extraction of data and analysis from several publically accessible databases and statistics, as well as from the LCA literature. Depending on the power generation technology, either plant-specific or region-specific average data have been used for creating the new power generation inventories representing specific geographies. Whenever possible, the parent–child relationship was used between global and local activities. All datasets include a specific technology level in order to support marginal mixes used in the consequential version of ecoinvent. The use of parameters, variables and mathematical relations enhances transparency. The article focuses on documentation of LCI data on the unlinked unit process level and presents direct emission data of the electricity-generating activities.

Results and discussion

Datasets for electricity production in 71 geographic regions (geographies) covering 50 countries are available in ecoinvent v3. The number of geographies exceeds the number of countries due to partitioning of power generation in the USA and Canada into several regions. All important technologies representing fossil, renewable and nuclear power are modelled for all geographies. The new inventory data show significant geography-specific variations: thermal power plant efficiencies, direct air pollutant emissions as well as annual yields of photovoltaic and wind power plants will have significant impacts on cumulative inventories. In general, the power plants operating in the 18 newly implemented countries (compared to ecoinvent v2) are on a lower technology level with lower efficiencies and higher emissions. The importance of local datasets is once more highlighted.

Conclusions

Inventories for average technology-specific electricity production in all globally important economies are now available with geography-specific technology datasets. This improved coverage of power generation representing 83 % of global electricity production in 2008 will increase the quality of and reduce uncertainties in LCA studies worldwide and contribute to a more accurate estimation of environmental burdens from global production chains. Future work on LCI of electricity production should focus on updates of the fuel chain and infrastructure datasets, on including new technologies as well as on refining of the local data.

     

Uncertainty analysis in LCA using precalculated aggregated datasets

Purpose

Some LCA software tools use precalculated aggregated datasets because they make LCA calculations much quicker. However, these datasets pose problems for uncertainty analysis. Even when aggregated dataset parameters are expressed as probability distributions, each dataset is sampled independently. This paper explores why independent sampling is incorrect and proposes two techniques to account for dependence in uncertainty analysis. The first is based on an analytical approach, while the other uses precalculated results sampled dependently.

Methods

The algorithm for generating arrays of dependently presampled aggregated inventories and their LCA scores is described. These arrays are used to calculate the correlation across all pairs of aggregated datasets in two ecoinvent LCI databases (2.2, 3.3 cutoff). The arrays are also used in the dependently presampled approach. The uncertainty of LCA results is calculated under different assumptions and using four different techniques and compared for two case studies: a simple water bottle LCA and an LCA of burger recipes.

Results and discussion

The meta-analysis of two LCI databases shows that there is no single correct approximation of correlation between aggregated datasets. The case studies show that the uncertainty of single-product LCA using aggregated datasets is usually underestimated when the correlation across datasets is ignored and that the magnitude of the underestimation is dependent on the system being analysed and the LCIA method chosen. Comparative LCA results show that independent sampling of aggregated datasets drastically overestimates the uncertainty of comparative metrics. The approach based on dependently presampled results yields results functionally identical to those obtained by Monte Carlo analysis using unit process datasets with a negligible computation time.

Conclusions

Independent sampling should not be used for comparative LCA. Moreover, the use of a one-size-fits-all correction factor to correct the calculated variability under independent sampling, as proposed elsewhere, is generally inadequate. The proposed approximate analytical approach is useful to estimate the importance of the covariance of aggregated datasets but not for comparative LCA. The approach based on dependently presampled results provides quick and correct results and has been implemented in EcodEX, a streamlined LCA software used by Nestlé. Dependently presampled results can be used for streamlined LCA software tools. Both presampling and analytical solutions require a preliminary one-time calculation of dependent samples for all aggregated datasets, which could be centrally done by database providers. The dependent presampling approach can be applied to other aspects of the LCA calculation chain.

     

Estimation of the size of error introduced into consequential models by using attributional background datasets

Purpose

A systematic comparison is made of attributional and consequential results for the same products using the same unit process database, thus isolating the effect of the two system models. An analysis of this nature has only recently been made possible due to the ecoinvent database version 3 providing an access to both unallocated and unlinked unit process datasets as well as both attributional and consequential models based on these datasets. The analysis is therefore limited to the system models provided by ecoinvent.

Methods

For both system models, the analysis was made on the life cycle inventory analysis (LCIA) results as published by ecoinvent (692 impact categories from different methods, for 11,650 product/activity combinations). The comparison was made on the absolute difference relative to the smallest absolute value.

Results and discussion

The comparison provides quantified results showing that the consequential modelling provides large differences in results when the unconstrained (marginal) suppliers have much more/less impact than the average, when analysing the by-products, and when analysing determining products from activities with important amounts of other coproducts.

Conclusions

The analysis confirms that for consequential studies, attributional background datasets are not appropriate as a substitute for consequential background. The overall error will of course depend on the extent to which attributional modelling is used as part of the overall system model. While the identified causes of differences between the attributional and consequential models are of general nature, the identified sizes of the errors are specific to the way the two models are implemented in ecoinvent.

     

The EcoSpold 2 format—why a new format?

Purpose

This paper introduces the new EcoSpold data format for life cycle inventory (LCI).

Methods

A short historical retrospect on data formats in the life cycle assessment (LCA) field is given. The guiding principles for the revision and implementation are explained. Some technical basics of the data format are described, and changes to the previous data format are explained.

Results

The EcoSpold 2 data format caters for new requirements that have arisen in the LCA field in recent years.

Conclusions

The new data format is the basis for the Ecoinvent v3 database, but since it is an open data format, it is expected to be adopted by other LCI databases. Several new concepts used in the new EcoSpold 2 data format open the way for new possibilities for the LCA practitioners and to expand the application of the datasets in other fields beyond LCA (e.g., Material Flow Analysis, Energy Balancing).

     

Giving a scientific basis for uncertainty factors used in global life cycle inventory databases: an algorithm to update factors using new information

Purpose

Life cycle inventory (LCI) databases provide generic data on exchange values associated with unit processes. The “ecoinvent” LCI database estimates the uncertainty of all exchange values through the application of the so-called pedigree approach. In the first release of the database, the used uncertainty factors were based on experts’ judgments. In 2013, Ciroth et al. derived empirically based factors. These, however, assumed that the same uncertainty factors could be used for all industrial sectors and fell short of providing basic uncertainty factors. The work presented here aims to overcome these limitations.

Methods

The proposed methodological framework is based on the assessment of more than 60 data sources (23,200 data points) and the use of Bayesian inference. Using Bayesian inference allows an update of uncertainty factors by systematically combining experts’ judgments and other information we already have about the uncertainty factors with new data.

Results and discussion

The implementation of the methodology over the data sources results in the definition of new uncertainty factors for all additional uncertainty indicators and for some specific industrial sectors. It also results in the definition of some basic uncertainty factors. In general, the factors obtained are higher than the ones obtained in previous work, which suggests that the experts had initially underestimated uncertainty. Furthermore, the presented methodology can be applied to update uncertainty factors as new data become available.

Conclusions

In practice, these uncertainty factors can systematically be incorporated in LCI databases as estimates of exchange value uncertainty where more formal uncertainty information is not available. The use of Bayesian inference is applied here to update uncertainty factors but can also be used in other life cycle assessment developments in order to improve experts’ judgments or to update parameter values when new data can be accessed.

     

The application of the pedigree approach to the distributions foreseen in ecoinvent v3

Purpose

Data used in life cycle inventories are uncertain (Ciroth et al. Int J Life Cycle Assess 9(4):216–226, 2004). The ecoinvent LCI database considers uncertainty on exchange values. The default approach applied to quantify uncertainty in ecoinvent is a semi-quantitative approach based on the use of a pedigree matrix; it considers two types of uncertainties: the basic uncertainty (the epistemic error) and the additional uncertainty (the uncertainty due to using imperfect data). This approach as implemented in ecoinvent v2 has several weaknesses or limitations, one being that uncertainty is always considered as following a lognormal distribution. The aim of this paper is to show how ecoinvent v3 will apply this approach to all types of distributions allowed by the ecoSpold v2 data format.

Methods

A new methodology was developed to apply the semi-quantitative approach to distributions other than the lognormal. This methodology and the consequent formulas were based on (1) how the basic and the additional uncertainties are combined for the lognormal distribution and on (2) the links between the lognormal and the normal distributions. These two points are summarized in four principles. In order to test the robustness of the proposed approach, the resulting parameters for all probability density functions (PDFs) are tested with those obtained through a Monte Carlo simulation. This comparison will validate the proposed approach.

Results and discussion

In order to combine the basic and the additional uncertainties for the considered distributions, the coefficient of variation (CV) is used as a relative measure of dispersion. Formulas to express the definition parameters for each distribution modeling a flow with its total uncertainty are given. The obtained results are illustrated with default values; they agree with the results obtained through the Monte Carlo simulation. Some limitations of the proposed approach are cited.

Conclusions

Providing formulas to apply the semi-quantitative pedigree approach to distributions other than the lognormal will allow the life cycle assessment (LCA) practitioner to select the appropriate distribution to model a datum with its total uncertainty. These data variability definition technique can be applied on all flow exchanges and also on parameters which play an important role in ecoinvent v3.

     

The ecoinvent database version 3 (part II): analyzing LCA results and comparison to version 2

Purpose

Version 3 of ecoinvent includes more data, new modeling principles, and, for the first time, several system models: the “Allocation, cut-off by classification” (Cut-off) system model, which replicates the modeling principles of version 2, and two newly introduced models called “Allocation at the point of substitution” (APOS) and “Consequential” (Wernet et al. 2016). The aim of this paper is to analyze and explain the differences in life cycle impact assessment (LCIA) results of the v3.1 Cut-off system model in comparison to v2.2 as well as the APOS and Consequential system models.

Methods

In order to do this, functionally equivalent datasets were matched across database versions and LCIA results compared to each other. In addition, the contribution of specific sectors was analyzed. The importance of new and updated data as well as new modeling principles is illustrated through examples.

Results and discussion

Differences were observed in between all database versions using the impact assessment methods Global Warming Potential (GWP100a), ReCiPe Endpoint (H/A), and Ecological Scarcity 2006 (ES’06). The highest differences were found for the comparison of the v3.1 Cut-off and v2.2. At average, LCIA results increased by 6, 8, and 17 % and showed a median dataset deviation of 13, 13, and 21 % for GWP, ReCiPe, and ES’06, respectively. These changes are due to the simultaneous update and addition of new data as well as through the introduction of global coverage and spatially consistent linking of activities throughout the database. As a consequence, supply chains are now globally better represented than in version 2 and lead, e.g., in the electricity sector, to more realistic life cycle inventory (LCI) background data. LCIA results of the Cut-off and APOS models are similar and differ mainly for recycling materials and wastes. In contrast, LCIA results of the Consequential version differ notably from the attributional system models, which is to be expected due to fundamentally different modeling principles. The use of marginal instead of average suppliers in markets, i.e., consumption mixes, is the main driver for result differences.

Conclusions

LCIA results continue to change as LCI databases evolve, which is confirmed by a historical comparison of v1.3 and v2.2. Version 3 features more up-to-date background data as well as global supply chains and should, therefore, be used instead of previous versions. Continuous efforts will be required to decrease the contribution of Rest-of-the-World (RoW) productions and thereby improve the global coverage of supply chains.

     

Attributional and consequential LCA in the ILCD handbook

Purpose

This discussion article aims to highlight two problematic aspects in the International Reference Life Cycle Data System (ILCD) Handbook: its guidance to the choice between attributional and consequential modeling and to the choice between average and marginal data as input to the life cycle inventory (LCI) analysis.

Methods

We analyze the ILCD guidance by comparing different statements in the handbook with each other and with previous research in this area.

Results and discussion

We find that the ILCD handbook is internally inconsistent when it comes to recommendations on how to choose between attributional and consequential modeling. We also find that the handbook is inconsistent with much of previous research in this matter, and also in the recommendations on how to choose between average and marginal data in the LCI.

Conclusions

Because of the inconsistencies in the ILCD handbook, we recommend that the handbook be revised.

     

What distribution function do life cycle inventories follow?

Purpose

Life cycle inventory (LCI) results are often assumed to follow a lognormal distribution, while a systematic study that identifies the distribution function that best describes LCIs has been lacking. This paper aims to find the distribution function that best describes LCIs using Ecoinvent v3.1 database using a statistical approach, called overlapping coefficient analysis.

Methods

Monte Carlo simulation is applied to characterize the distribution of aggregate LCIs. One thousand times of simulated LCI results are generated based on the unit process-level parametric uncertainty information, from each of which 1000 randomly chosen data points are extracted. The 1 million data points extracted undergo statistical analyses including Shapiro-Wilk normality test and the overlapping coefficient analysis. The overlapping coefficient is a measure used to determine the shared area between the distribution of the simulated LCI results and three possible distribution functions that can potentially be used to describe them including lognormal, gamma, and Weibull distributions.

Results and discussion

Shapiro-Wilk normality test for 1000 samples shows that average p value of log-transformed LCI results is 0.18 at 95 % confidence level, accepting the null hypothesis that LCI results are lognormally distributed. The overlapping coefficient analysis shows that lognormal distribution best describes the distribution of LCI results. The average of overlapping coefficient (OVL) for lognormal distribution is 95 %, while that for gamma and Weibull distributions are 92 and 86 %, respectively.

Conclusions

This study represents the first attempt to calculate the stochastic distributions of the aggregate LCIs covering the entire Ecoinvent 3.1 database. This study empirically shows that LCIs of Ecoinvent 3.1 database indeed follow a lognormal distribution. This finding can facilitate more efficient storage and use of uncertainty information in LCIs and can reduce the demand for computational power to run Monte Carlo simulation, which currently relies on unit process-level uncertainty data.

     

Bridge processes: a solution for LCI datasets independent of background databases

Introduction

New platforms are emerging that enable more data providers to publish life cycle inventory data.

Background

Providing datasets that are not complete LCA models results in fragments that are difficult for practitioners to integrate and use for LCA modeling. Additionally, when proxies are used to provide a technosphere input to a process that was not originally intended by the process authors, in most LCA software, this requires modifying the original process.

Results

The use of a bridge process, which is a process created to link two existing processes, is proposed as a solution.

Discussion

Benefits to bridge processes include increasing model transparency, facilitating dataset sharing and integration without compromising original dataset integrity and independence, providing a structure with which to make the data quality associated with process linkages explicit, and increasing model flexibility in the case that multiple bridges are provided. A drawback is that they add additional processes to existing LCA models which will increase their size.

Conclusions

Bridge processes can be an enabler in allowing users to integrate new datasets without modifying them to link to background databases or other processes they have available. They may not be the ideal long-term solution but provide a solution that works within the existing LCA data model.

     

NS-kNN: a modified <Emphasis Type="Italic">k</Emphasis>-nearest neighbors approach for imputing metabolomics data

Introduction

A common problem in metabolomics data analysis is the existence of a substantial number of missing values, which can complicate, bias, or even prevent certain downstream analyses. One of the most widely-used solutions to this problem is imputation of missing values using a k-nearest neighbors (kNN) algorithm to estimate missing metabolite abundances. kNN implicitly assumes that missing values are uniformly distributed at random in the dataset, but this is typically not true in metabolomics, where many values are missing because they are below the limit of detection of the analytical instrumentation.

Objectives

Here, we explore the impact of nonuniformly distributed missing values (missing not at random, or MNAR) on imputation performance. We present a new model for generating synthetic missing data and a new algorithm, No-Skip kNN (NS-kNN), that accounts for MNAR values to provide more accurate imputations.

Methods

We compare the imputation errors of the original kNN algorithm using two distance metrics, NS-kNN, and a recently developed algorithm KNN-TN, when applied to multiple experimental datasets with different types and levels of missing data.

Results

Our results show that NS-kNN typically outperforms kNN when at least 20–30% of missing values in a dataset are MNAR. NS-kNN also has lower imputation errors than KNN-TN on realistic datasets when at least 50% of missing values are MNAR.

Conclusion

Accounting for the nonuniform distribution of missing values in metabolomics data can significantly improve the results of imputation algorithms. The NS-kNN method imputes missing metabolomics data more accurately than existing kNN-based approaches when used on realistic datasets.

     

Network-based logistic regression integration method for biomarker identification

Background

Many mathematical and statistical models and algorithms have been proposed to do biomarker identification in recent years. However, the biomarkers inferred from different datasets suffer a lack of reproducibilities due to the heterogeneity of the data generated from different platforms or laboratories. This motivates us to develop robust biomarker identification methods by integrating multiple datasets.

Methods

In this paper, we developed an integrative method for classification based on logistic regression. Different constant terms are set in the logistic regression model to measure the heterogeneity of the samples. By minimizing the differences of the constant terms within the same dataset, both the homogeneity within the same dataset and the heterogeneity in multiple datasets can be kept. The model is formulated as an optimization problem with a network penalty measuring the differences of the constant terms. The L₁ penalty, elastic penalty and network related penalties are added to the objective function for the biomarker discovery purpose. Algorithms based on proximal Newton method are proposed to solve the optimization problem.

Results

We first applied the proposed method to the simulated datasets. Both the AUC of the prediction and the biomarker identification accuracy are improved. We then applied the method to two breast cancer gene expression datasets. By integrating both datasets, the prediction AUC is improved over directly merging the datasets and MetaLasso. And it’s comparable to the best AUC when doing biomarker identification in an individual dataset. The identified biomarkers using network related penalty for variables were further analyzed. Meaningful subnetworks enriched by breast cancer were identified.

Conclusion

A network-based integrative logistic regression model is proposed in the paper. It improves both the prediction and biomarker identification accuracy.

     

Thai national life cycle inventory readiness for product environmental footprint

Purpose

In the near future, the products of Thai industries and companies mainly producing parts and products for export to the European Union (EU) will require the Product Environmental Footprint (PEF) to assess the environmental performance and resource efficiency of products by using a life cycle perspective. The potential generic (often used interchangeably with background data) data have to be modified and improved for mandatory use in the product-specific and country-specific PEF database.

Methods

PEF is used as a tool for assessing the environmental burden of products and services for export to the EU. It requires both specific data from primary sources and generic data to fulfill assessment requirement. Accordingly, the Thai national life cycle inventory (LCI) database plays a key role in generic data that was used to evaluate the environmental performance of products. This paper presents the perspective of Thai data readiness for PEF in which the quality of LCI is the main issue of concern. The current situation of the Thai national LCI database was reviewed. Then, the gaps of data were addressed, and the gaps were also filled. Non-representative data and untreated waste are the selected issues that were presented in this paper.

Results and discussion

Many gaps were revealed for the Thai national LCI database because this database was developed based on ISO 14040/44, which may not be compliant with the PEF guide. The issues that have been selected for improvement are non-representative data and untreated waste because these gaps can offer inaccuracy concerning the environmental burden of products potentially leading to the reliability of products for export to the EU. However, the Thai national LCI database has not achieved the data quality aspects of the PEF, continuously improving the quality of data to meet the requirements of the PEF.

Conclusions

The lessons learned from the real-world situation of data quality development based on PEF requirements were extracted. The practical procedure and recommendations were transparent for drivers and researchers who would like to start with data quality issues and prepare for the EU single market.

     

Empirical study of seven data mining algorithms on different characteristics of datasets for biomedical classification applications

Background

Various kinds of data mining algorithms are continuously raised with the development of related disciplines. The applicable scopes and their performances of these algorithms are different. Hence, finding a suitable algorithm for a dataset is becoming an important emphasis for biomedical researchers to solve practical problems promptly.

Methods

In this paper, seven kinds of sophisticated active algorithms, namely, C4.5, support vector machine, AdaBoost, k-nearest neighbor, naïve Bayes, random forest, and logistic regression, were selected as the research objects. The seven algorithms were applied to the 12 top-click UCI public datasets with the task of classification, and their performances were compared through induction and analysis. The sample size, number of attributes, number of missing values, and the sample size of each class, correlation coefficients between variables, class entropy of task variable, and the ratio of the sample size of the largest class to the least class were calculated to character the 12 research datasets.

Results

The two ensemble algorithms reach high accuracy of classification on most datasets. Moreover, random forest performs better than AdaBoost on the unbalanced dataset of the multi-class task. Simple algorithms, such as the naïve Bayes and logistic regression model are suitable for a small dataset with high correlation between the task and other non-task attribute variables. K-nearest neighbor and C4.5 decision tree algorithms perform well on binary- and multi-class task datasets. Support vector machine is more adept on the balanced small dataset of the binary-class task.

Conclusions

No algorithm can maintain the best performance in all datasets. The applicability of the seven data mining algorithms on the datasets with different characteristics was summarized to provide a reference for biomedical researchers or beginners in different fields.