Data-driven Decision Support Systems Research Articles

THU163 Traffic Light Alerting System For Adherence Risk Management To Improve Catch-Up Growth In Patients Receiving Recombinant Human Growth Hormone Therapy

Abstract Disclosure: S. Loche: Advisory Board Member; Self; SL has received advisory board fees from the healthcare business of Merck KGaA, Darmstadt, Germany. Consulting Fee; Self; SL has received consultancy fees from the healthcare business of Merck KGaA, Darmstadt, Germany. Speaker; Self; SL has received lecture fees from the healthcare business of Merck KGaA, Darmstadt, Germany. P. van Dommelen: Consulting Fee; Self; PvD has a consultancy agreement with the healthcare business of Merck KGaA, Darmstadt, Germany. E. Koledova: Employee; Self; EK is an employee of the healthcare business of Merck KGaA, Darmstadt, Germany. Stock Owner; Self; EK holds shares in the healthcare business of Merck KGaA, Darmstadt, Germany. Background: In our previous study, a data-driven clinical decision support system based on “traffic light” visualizations for adherence risk management in patients receiving recombinant human growth hormone (r-hGH) treatment was developed.1 Two “traffic lights” were most promising: mean adherence and standard deviation (SD) of hours to next injection (administering injections at around the same time each day). We aimed to study the effect of these two adherence-based traffic lights on catch-up growth in the first year of treatment in patients with growth hormone deficiency (GHD) and those born small for gestational age (SGA). Data and Methods: Height and adherence data were extracted from the easypod™ connect ecosystem and taken from the easypod™ connect observational study.2 Patients with: height standard deviation scores (HSDS) of ≤-2 at treatment start; age 2-15 years at treatment start; and ≥1 measurement and adherence data available in the first year of treatment, were selected. Mean adherence was classified as high (≥85%), intermediate (>56%-<85%), or low (≤56%); high (≥15.9 [P95]), intermediate (≥10.1 [P85]), or low (<10.1) were used for SD of hours to next injection. Linear regression analysis was performed with ΔHSDS between treatment start and first year of treatment as dependent variables and the adherence-based values as independent variables, all adjusted for age and HSDS at treatment start. Results: In total, data for 1,045 patients (776 and 269 with GHD and those born SGA, respectively) were available. Only some of the adherence-based categories had a significant and independent effect on ΔHSDS (P=0.004). Adjusted for age and HSDS at treatment start, patients with high/intermediate mean adherence had, on average, a +0.15 SD higher ΔHSDS (P=0.004), while this was +0.09 SD for patients with a low SD of hours to the next injection (P=0.003). Therefore, patients with both high/intermediate mean adherence and a low SD of hours to the next injection had, on average, a +0.24 SD higher ΔHSDS compared with patients with both low mean adherence and a high SD of hours to the next injection. Conclusions: Our research shows that good adherence and administering injections around the same time each day play an essential role in optimizing catch-up growth. Adherence-based traffic lights can alert clinicians to have discussions with patients/caregivers to mitigate the risk of sub-optimal adherence and, consequently, improve catch-up growth.

A data-driven decision support system for service completion prediction in last mile logistics

A data-driven decision support system for service completion prediction in last mile logistics

Temporal characterization of acute pain and toxicity kinetics during radiation therapy for head and neck cancer. A retrospective study.

Pain during Radiation Therapy (RT) for oral cavity/oropharyngeal cancer (OC/OPC) is a clinical challenge due to its multifactorial etiology and variable management. The objective of this study was to define complex pain profiles through temporal characterization of pain descriptors, physiologic state, and RT-induced toxicities for pain trajectories understanding. Using an electronic health record registry, 351 OC/OPC patients treated with RT from 2013 to 2021 were included. Weekly numeric scale pain scores, pain descriptors, vital signs, physician-reported toxicities, and analgesics were analyzed using linear mixed effect models and Spearman's correlation. Area under the pain curve (AUCpain) was calculated to measure pain burden over time. Median pain scores increased from 0 during the weekly visit (WSV)-1 to 5 during WSV-7. By WSV-7, 60% and 74% of patients reported mouth and throat pain, respectively, with a median pain score of 5. Soreness and burning pain peaked during WSV-6/7 (51%). Median AUCpain was 16% (IQR (9.3-23)), and AUCpain significantly varied based on gender, tumor site, surgery, drug use history, and pre-RT pain. A temporal increase in mucositis and dermatitis, declining mean bodyweight (-7.1%; P < 0.001) and mean arterial pressure (MAP) 6.8 mmHg; P < 0.001 were detected. Pulse rate was positively associated while weight and MAP were negatively associated with pain over time (P < 0.001). This study provides insight on in-depth characterization and associations between dynamic pain, physiologic, and toxicity kinetics. Our findings support further needs of optimized pain control through temporal data-driven clinical decision support systems for acute pain management.

Design an efficient data driven decision support system to predict flooding by analysing heterogeneous and multiple data sources using Data Lake

Floods are the most common natural disaster in several countries throughout the world. Flooding has a major impact on people's lives and livelihoods. The impact of flood disasters on human lives can be mitigated by developing effective flood forecasting and prediction models. The majority of flood prediction models do not take all flood-causing factors into account when they are designed. It is difficult to collect and handle some of these flood-causing variables since they are heterogeneous in nature. This paper presents a new big data architecture called Data Lake, which can ingest and store all important flood-causing heterogeneous data sources in their raw format for machine learning model creation. The statistical relevance of important flood producing factors on flood prediction outcome is determined utilizing inferential statistical approaches. The outcome of this research is to create flood warning systems that can alert the public and government officials so that they can make decisions in the event of a severe flood, reducing socioeconomic loss.•Flood causing factors are from heterogeneous sources, so there is no big data architecture for handling variety of data sources.•To provide data architectural solution using data lake for collecting and analysing heterogeneous flood causing factors.•Uses inferential statistical approach to determine importance of different flood causing factors in design of efficient flood prediction models.

A data-driven intelligent decision support system that combines predictive and prescriptive analytics for the design of new textile fabrics

In this paper, we propose an Intelligent Decision Support System (IDSS) for the design of new textile fabrics. The IDSS uses predictive analytics to estimate fabric properties (e.g., elasticity) and composition values (% cotton) and then prescriptive techniques to optimize the fabric design inputs that feed the predictive models (e.g., types of yarns used). Using thousands of data records from a Portuguese textile company, we compared two distinct Machine Learning (ML) predictive approaches: Single-Target Regression (STR), via an Automated ML (AutoML) tool, and Multi-target Regression, via a deep learning Artificial Neural Network. For the prescriptive analytics, we compared two Evolutionary Multi-objective Optimization (EMO) methods (NSGA-II and R-NSGA-II) when optimizing 100 new fabrics, aiming to simultaneously minimize the physical property predictive error and the distance of the optimized values when compared with the learned input space. The two EMO methods were applied to design of 100 new fabrics. Overall, the STR approach provided the best results for both prediction tasks, with Normalized Mean Absolute Error values that range from 4% (weft elasticity) to 11% (pilling) in terms of the fabric properties and a textile composition classification accuracy of 87% when adopting a small tolerance of 0.01 for predicting the percentages of six types of fibers (e.g., cotton). As for the prescriptive results, they favored the R-NSGA-II EMO method, which tends to select Pareto curves that are associated with an average 11% predictive error and 16% distance.

Data-driven decision support system for degrading assets and its application under the perspective of a railway component

Transportation plays a crucial role in facilitating economic and social interactions, but also significantly contributes to global pollution. To address this sustainability challenge, smart use of resources and transport modes is needed. To contribute to more efficient mobility in worldwide transportation, railways should have better maintenance. Within railway vehicles, the wheelset component is essential for the safety of the train, preventing derailment. However, degradation resulting from the wheel-rail contact during operation may compromise the proper functioning of the components, requiring maintenance to be carried out. Nowadays, the accumulation in the last years of information and the technological ease to generate large amounts of data in inspection and maintenance activities in railway companies allow for implementing robust and reliable data-driven approaches to improve the maintenance of the components. With this scenario in the background, this paper proposes a data-driven decision support system for maintenance actions in degrading assets. Predictive degradation models of wear and damage are developed. The uncertainty of degradation data obtained in measurement condition monitoring activities is quantified and used to evaluate the impact on optimal maintenance strategies. Results obtained indicate that optimized maintenance actions can extend the lifespan of the wheelset and rail components and reduce their long-term maintenance costs. Outputs from the decision support system suggest better maintenance practices to be performed in railway companies.

A data-driven decision support system for sustainable supplier evaluation in the Industry 5.0 era: A case study for medical equipment manufacturing

Effective supplier management is critical for an enterprise’s success, as supplier procurement accounts for up to approximately 70% to 80% of total manufacturing costs. Correct supplier selection can ensure the competitiveness of the enterprise and contribute to supply chain integration and innovation. In recent years, supplier evaluation frameworks based on Industry 4.0 concepts have contributed to the development of the industry. However, the novel concepts of Industry 5.0 require examination from a people-oriented and sustainable perspective. Unfortunately, at the present time, supplier evaluation frameworks based on Industry 5.0 are lacking. Therefore, the primary task of this study is to develop a novel and comprehensive supplier evaluation framework for the Industry 5.0 era. This study proposes a data-driven decision support system to execute the supplier evaluation process. First, variable precision- dominance-based rough set approach (VC-DRSA) is applied to extract the core criteria, to remove the noise factors and to generate decision rules for the decision-makers’ reference. Second, the criterion importance through intercriteria correlation (CRITIC) approach is adopted to obtain the dependency weights of the core criteria and their ranking. Finally, a modified classifiable technique for order preference by similarity to ideal solution (CTOPSIS) is used to integrate the final performance values of suppliers when new alternative suppliers are added. The research concept is in line with the conception of data-driven decision support in business intelligence and does not rely on the subjective judgments and opinions of experts. Data provided by a multinational medical equipment manufacturer are used as an example to demonstrate the proposed model. VC-DRSA retains nine core criteria from the original twenty criteria, which greatly reduces the labor and cost of supplier audits. In addition, the CRITIC results show that digital transformation, real-time information sharing, and organizational culture transformation are the three main factors affecting the development of enterprises towards Industry 5.0. The results show that CTOPSIS can be used to quickly assess the ratings of new alternative suppliers are listed.

Pay-for-performance programs effectiveness in healthcare: the case of the end-stage renal disease quality incentive program.

This paper focuses on Medicare's End-Stage Renal Disease Quality Incentive Program (QIP). QIP aims to promote high-quality services in outpatient dialysis facilities by tying their payments to their performance on pre-specified quality measures. In this paper, employing principal-agent theory, we examine the effectiveness of QIP by exploring the changes in various clinical/operational measures when they become a part of the program as a performance measure. We study five QIP quality measures; two are operational: hospitalization and readmission. And three others are clinical: blood transfusion, hypercalcemia, and dialysis adequacy. Overall, we observe a significant improvement in all QIP quality measures after being included in the program, except for readmission. We recommend adjusting the weight and redesigning the readmission measure for Medicare to incentivize providers to reduce readmission. We also discuss establishing care coordination and employing data-driven clinical decision support systems as opportunities for dialysis facilities to improve the care delivery process.

Assessment of Micronutrient Content and Strategies for Improvement of Micronutrient Status in the Soils of Bhandara Tehsil, Maharashtra, India

The study was undertaken to determine the available micro-nutrient status of the Bhandara tehsil of Maharashtra with the help of GPS-based one hundred-and-five soil samples during the period of 2021-2022 from paddy growing fields. Samples were analyzed at the Soil Science and Agricultural Chemistry Section, College of Agriculture, Nagpur. The results revealed that soils were clay loam to sandy clay loam in texture. The study area was slightly acidic to slightly alkaline and non-saline in reaction. The mean values of 0.42 mg kg-1, 5.44 mg kg-1, 6.39 mg kg-1, 0.98 mg kg-1, and 0.90 mg kg-1 for Zn, Fe, Mn, Cu and B respectively, were recorded. The soil nutrient index values showed that available Zn was low; available Fe was moderate; available Mn and B were moderately high; and available Cu was high. The farmers of the study area mostly adopt improper nutrient management practices which might have resulted in poor fertility status in the area. Taking these things into consideration, a data-driven Decision Support System (DSS) on micronutrient management in soil has been given. To address this problem, farmers of Bhandara tehsil may use micronutrient fertilizers or implement soil management practices that promote the uptake and availability of micronutrients to plants.

Towards an Integrated Approach for Preserving Data Utility, Privacy and Fairness

Data reusability has become a distinct characteristic of scientific, commercial, and administrative practices nowadays. However, an unlimited and careless reuse of data may lead to privacy breaches and unfair impacts on individuals and vulnerable groups. Data content adaption is a key aspect of preserving data privacy and fairness. Often, such content adaption affects data utility adversely. Further, the interaction between privacy protection and fairness protection can be subject to making trade-offs because mitigating privacy risks may adversely affect detecting unfairness and vice versa. Therefore, there is a need for research on understanding the interactions between data utility, privacy and fairness. To this end, in this contribution, we use concepts from causal reasoning and argue for adopting an integrated view on data content adaption for data driven decision support systems. This asks for considering the operation context wholistically. By means of two cases, we illustrate that, in some situations, local data content adaption may lead to low data quality and utility. An integrated wholistic approach, however, may result in reuse of the original data (i.e., without content adaption, thus in higher data utilization) without adversely affecting privacy and fairness. We discuss some implications of this approach and sketch a few directions for future research.

A decision support model to evaluate liveability in the context of urban vibrancy

Liveability can be accepted as an umbrella term covering all the factors that make a place to live. We recognize the versatility of urban liveability and focus on the vibrancy aspect. Regarding the literature, we compile variables affecting urban liveability under the economic, image, and use value of place. This article aims to present a data-driven decision support system to evaluate different dimensions of vibrancy-focused liveability. We adopt a knowledge discovery process to handle the complexity of the liveability concept. This study develops a conditional-based relationship network of vibrancy parameters through the Bayesian Belief Network (BBN). Then, we assess the BBN's correlations with statistics and causal relations with the survey in this study.These results mostly agree with the findings of the relevant literature. The economic value results show that the high density, diversity and accessibility add a premium to the land value of properties. The use value results also demonstrate that the diversity and density of activities, cultural attributes, and high accessibility support place attractiveness. The selected streetscape variables improve image value, except for building enclosure and condition. The study has the potential for urban planners to vitalize neighborhoods by considering urban activities and urban physical attributes.

A data-driven decision support system with smart packaging in grocery store supply chains during outbreaks

The unexpected emergence of the COVID-19 pandemic has changed how grocery shopping is done. The grocery retail stores need to ensure hygiene, quality, and safety concerns in-store shopping by providing “no-touch” smart packaging solutions for agri-food products. The benefit of smart packaging is to inform consumers about the freshness level of a packaged product without having direct contact. This paper proposes a data-driven decision support system that uses smart packaging as a smart product-service system to manage the sustainable grocery store supply chain during outbreaks to prevent food waste. The proposed model dynamically updates the price of a packaged perishable product depending on freshness level while reducing food waste and the number of rejected customers and maximising profit by increasing the inventory turnover rate of grocery stores. The model was tested on a hypothetical but realistic case study of a single product. The results of this study showed that stock capacities, freshness discount rate, freshness period, and quantity discounts significantly affect the performance of a grocery store supply chain during outbreaks.

A Platform Approach to Smart Farm Information Processing

With the rapid growth of population and the increasing demand for food worldwide, improving productivity in farming procedures is essential. Smart farming is a concept that emphasizes the use of modern technologies such as the Internet of Things (IoT) and artificial intelligence (AI) to enhance productivity in farming practices. In a smart farming scenario, large amounts of data are collected from diverse sources such as wireless sensor networks, network-connected weather stations, monitoring cameras, and smartphones. These data are valuable resources to be used in data-driven services and decision support systems (DSS) in farming applications. However, one of the major challenges with these large amounts of agriculture data is their immense diversity in terms of format and meaning. Moreover, the different services and technologies in a smart farming ecosystem have limited capability to work together due to the lack of standardized practices for data and system integration. These issues create a significant challenge in cooperative service provision, data and technology integration, and data-sharing practices. To address these issues, in this paper, we propose the platform approach, a design approach intended to guide building effective, reliable, and robust smart farming systems. The proposed platform approach considers six requirements for seamless integration, processing, and use of farm data. These requirements in a smart farming platform include interoperability, reliability, scalability, real-time data processing, end-to-end security and privacy, and standardized regulations and policies. A smart farming platform that considers these requirements leads to increased productivity, profitability, and performance of connected smart farms. In this paper, we aim at introducing the platform approach concept for smart farming and reviewing the requirements for this approach.

ENSURE: A Deep Learning Approach for Enhancing Situational Awareness in Surveillance Applications With Ubiquitous High-Dimensional Sensing

Judicious processing of <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">High-Dimensional Sensing (HDS)</i> data streams is often the first step in a data-driven decision-support system. However, how one goes about processing large amounts of incoming HDS data streams for enhancing <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">situational awareness (SA)</i> is a non-trivial, and often challenging task due to the interwoven nature of operational environments, especially in complex surveillance applications ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">e.g.,</i> predictive policing, urban reconnaissance). Toward addressing this challenge, a novel data processing framework called <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">ENSURE (ENhanced Situational Understanding with Ubiquitous-sensing REsources)</i> is presented. In particular, by using <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">density-based clustering</i> in tandem with variable length sequence decoding methods, ubiquitous HDS data streams are clustered and then processed for early identification of ‘events’ and their sequences of interest to a (surveillance) task at hand. Then, by leveraging notions in <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">data association,</i> ENSURE ‘tracks’ the occurrence of such event sequences and their evolution over time to ‘perceive’ and ‘comprehend’ what’s happening in the operational environment. Therefore, ENSURE can be interpreted as an <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">information fusion</i> approach for HDS systems for object recognition, classification, tracking, and decision-making tasks. Use of ENSURE is illustrated via a simulated experiment to provide insights into its behavior, performance and sensitivity to parameters.

Data-driven decision support system for building stocks energy retrofit policy

In most European countries, residential assets account for as much as 85% of the building stock floor area and are, on average, very outdated and energy inefficient. Moreover, the European Commission published the EU Green Deal invigorating higher retrofit of private and public buildings. Nowadays, public authorities collect extensive datasets to analyze the existing building stock; however, the complex and diverse scenario makes the definition of retrofit policies cumbersome. The biggest hurdle is often linked to the high cost of acquiring information. The presented research tries to overcome these issues by introducing a decision support system for retrofit policymaking from low-cost data-driven approaches. The method is based on: i) clustering techniques to divide building assets into groups with similar characteristics and energy consumption, and ii) Montecarlo simulation to compute each cluster's energy savings based on different retrofit scenarios. The proposed method has been successfully applied to an extensive portfolio of residential assets in Lombardy Region in Italy, called the CENED database, with over one million assets. As a result, the introduced method defines the optimum retrofit scenario with a low cost of information (e.g., without expensive surveys to gather data on existing assets' characteristics and performance indicators) and determines the number of assets to be retrofitted along with the expected energy savings. This data-driven approach can be easily updated given new renovations and status changes in the built environment, making it useable for the long term or in different regions. To summarize, data-driven solutions are now required to accomplish the European Union's decarbonization ambitions, and the proposed method helps decision-makers choose better energy retrofit policies for the built environment.

A data-driven stochastic decision support system to investment portfolio problem under uncertainty

A data-driven stochastic decision support system to investment portfolio problem under uncertainty

Prediction of clinical trial enrollment rates.

Clinical trials represent a critical milestone of translational and clinical sciences. However, poor recruitment to clinical trials has been a long standing problem affecting institutions all over the world. One way to reduce the cost incurred by insufficient enrollment is to minimize initiating trials that are most likely to fall short of their enrollment goal. Hence, the ability to predict which proposed trials will meet enrollment goals prior to the start of the trial is highly beneficial. In the current study, we leveraged a data set extracted from ClinicalTrials.gov that consists of 46,724 U.S. based clinical trials from 1990 to 2020. We constructed 4,636 candidate predictors based on data collected by ClinicalTrials.gov and external sources for enrollment rate prediction using various state-of-the-art machine learning methods. Taking advantage of a nested time series cross-validation design, our models resulted in good predictive performance that is generalizable to future data and stable over time. Moreover, information content analysis revealed the study design related features to be the most informative feature type regarding enrollment. Compared to the performance of models built with all features, the performance of models built with study design related features is only marginally worse (AUC = 0.78 ± 0.03 vs. AUC = 0.76 ± 0.02). The results presented can form the basis for data-driven decision support systems to assess whether proposed clinical trials would likely meet their enrollment goal.

Towards Data Driven Spatio-Temporal Threshold Identification Based on Cost Effective Public Health Information Management Framework

Appropriate public health action comes from data driven decision support systems. While sophisticated health information exchange framework may be costly in developing countries, the health care delivery system in place may provide a promising infrastructure that spans all parts in a region. Therefore, while digital and non digital data is constantly being generated from variety of sources including public and private health sectors, the health care delivery systems remain the primary and most fundamental source for data on population health status. For low and middle income countries with minimum digitization and resource constraints, traditional existing health care delivery system can be taken advantage of, for efficient production and timely transmission and utilisation of data to identify thresholds for disease outbreak in order to improve health status and health system performance. Due to lack of appropriate universally agreed criteria for threshold, defining local thresholds for infectious disease is not only crucial but also more appropriate. In this paper, we present a low cost data driven framework called Health Data Driven Framework (HDDF) through which data generated at health care facilities may be used for threshold detection and alarm generation. We also identify a localized method based on spatio-temporal mining of available data for appropriate threshold identification.

A Deep Learning Framework for COVID-19 Diagnosis from Computed Tomography

The outbreak of novel Coronavirus COVID-19, an infectious disease caused by the SARS-CoV-2 virus, has caused an unprecedented medical, economic, and social emergency that requires data-driven intelligence and decision support systems to counter the subsequent pandemic. Data-driven models and intelligent systems can assist medical researchers and practitioners to identify symptoms of COVID-19 infection. Several solutions based on medical image processing have been proposed for this purpose. However, the most shortcoming of hand craft image processing systems is the lower provided performances. Hence, for the first time, the proposed solution uses a deep learning model that is applied to Computed Tomography (CT) images for the efficient extraction of COVID-19 features. Since there are few patients in the COVID-CT-Dataset, the Convolutional Neural Network (CNN) model cannot undergo further learned to enhance performances. Therefore, the proposed solution works as a pipeline framework involving two steps: (A) baseline classification is provided by a CNN model; (B) baseline results are re-ranked using distances to features vectors of CT image parts. A re-ranking framework is used as additional means of COVID-19 symptom identification. These steps exploit the diversity of different parts of CT images to enhance classification performance. Evaluations of the proposed solution are driven by real world data based on clinical findings in the form of COVID-CT-Dataset images. The results of the evaluation illustrate the streamlined efficiency and accuracy of the proposed solution to the image-based diagnosis of COVID-19 patients. Our findings support smart healthcare solutions–specifically addressing COVID-19 challenges–and provide guidelines to engineer and develop intelligent and autonomous systems.

An AI-enhanced Approach for optimizing life cycle costing of military logistic vehicles

Individual usage profile (aka true usage profile) of military logistic vehicles (MLV) is highly dependent on types of operations, duration and areas of use. Compared to non-military vehicles, the operation life is considerably longer, mileage tends to be several times lower and the characteristics of the application area, including environmental factors, is much more diverse. Transparent and realistic determination of MLV’s true usage profile facilitates, inter alia, optimal life cycle costing in particular operation costs. In practice, however, effects of non-usage, such as material degradation or damage during service of the vehicles are not explicitly recorded. Important aspects of assets management, such as identifying maintenance measures and logistic disposition calculations, are often conducted non-systematically and non-data driven but rather based on human experiential knowledge and manufacturer’s specifications. Hence, the modelling and predictive analysis of MLV’s true usage profile is subject to significant uncertainties. The body of knowledge in assets management does not consider the aforementioned problems. Therefore, lack of novel AI-enhanced approaches for modelling MLV’s true usage profile is evident. This paper introduces an integrative, data-driven approach for identifying the true usage profile and thus improving the life cycle costing of MLV. The proposed approach involves an AI-enhanced analysis of heterogeneous data sources, such as sensor-, vehicle control-, GPS-, and logistics data as well as textual logbooks. It enables a transparent and realistic identification of MLV’s true usage profiles, which facilitates accurate and timely calculation of life cycle costs under consideration of data quality properties. Finally yet importantly, classification of usage profiles enables a data-driven decision support system to provide i) proactive maintenance measures, and ultimately ii) to optimize distribution of vehicles and spare parts in decentralized warehouses.