In machine learning, hyperparameter tuning is strongly useful to improve model performance. In ou... more In machine learning, hyperparameter tuning is strongly useful to improve model performance. In our research, we concentrate our attention on classifying imbalanced data by cost-sensitive support vector machines. We propose a multi-objective approach that optimizes model’s hyper-parameters. The approach is devised for imbalanced data. Three SVM model’s performance measures are optimized. We present the algorithm in a basic version based on genetic algorithms, and as an improved version based on genetic algorithms combined with decision trees. We tested the basic and the improved approach on benchmark datasets either as serial and parallel version. The improved version strongly reduces the computational time needed for finding optimized hyper-parameters. The results empirically show that suitable evaluation measures should be used in assessing the classification performance of classification models with imbalanced data.
Springer Proceedings in Mathematics & Statistics, 2017
This work presents an innovative quantitative approach, based on an optimization model, to manage... more This work presents an innovative quantitative approach, based on an optimization model, to manage outpatient Day Service operations. The main objective of the research work is to maximize admitted patient flow in order to use available resources efficiently. The model is designed with the help of Answer Set Programming and implemented in Datalog with Disjunction (DLV) system. As significant case study, we consider the Rheumatology domain. Preliminary results are presented.
We propose a new set of clinical variables for a more accurate early prediction of safe decannula... more We propose a new set of clinical variables for a more accurate early prediction of safe decannulation in patients with severe acquired brain injury (ABI), during a post-acute rehabilitation course. Starting from the already validated DecaPreT scale, we tested the accuracy of new logistic regression models where the coefficients of the original predictors were reestimated. Patients with tracheostomy were retrospectively selected from the database of the neurorehabilitation unit at the S. Anna Institute of Crotone, Italy. New potential predictors of decannulation were screened from variables collected on admission during clinical examination, including (a) age at injury, (b) coma recovery scale-revised (CRS-r) scores, and c) length of ICU period. Of 273 patients with ABI (mean age 53.01 years; 34% female; median DecaPreT = 0.61), 61.5% were safely decannulated before discharge. In the validation phase, the linear logistic prediction model, created with the new multivariable predictors...
2018 Computing in Cardiology Conference (CinC), 2018
This work aims at developing and assessing a machine learning based Knowledge Discovery task for ... more This work aims at developing and assessing a machine learning based Knowledge Discovery task for risk prediction of major cardiovascular worsening events in chronic heart failure patients. Clinical data from 50patients with chronic heartfailure was analyzed. For each patient, personal data, different vital and clinical parameters and the presence of cardiovascular worsening events have been stored every three months per two years. We defined the Knowledge Discovery analysis as a predictive task stated as supervised binary classification problem. The class label was defined based on the occurrence or not of cardiovascular worsening events between two consecutive visits. To take into account the temporality of the worsening events, six different temporal weighting strategies, applied to the vital parameters, were tested. Several machine learning algorithms were applied for each strategy obtaining different predictive models. Models performance have been evaluated mainly in term of are...
Patient admission and surgery scheduling is a complex combinatorial optimization problem. It cons... more Patient admission and surgery scheduling is a complex combinatorial optimization problem. It consists on defining patient admission dates, assigning them to suitable rooms, and schedule surgeries accordingly to an existing master surgical schedule. This problem belongs to the class of NP-hard problems. In this paper, we firstly formulate an integer programming model for offline patient admissions, room assignments, and surgery scheduling; then apply a matheuristic that combines exact methods with rescheduling approaches. The matheuristic is evaluated using benchmark datasets. The experimental results improve those reported in the literature and show that the proposed method outperforms existing techniques of the state-of-the-arts.
This paper originates from the HeartDrive project, a platform of services for a more effective, e... more This paper originates from the HeartDrive project, a platform of services for a more effective, efficient and integrated management of heart failure and comorbidities. HeartDrive establishes a cooperative approach based on the concepts of continuity of care and extreme, patient oriented, customization of diagnostic, therapeutic and follow-up procedures. Definition and development of evidence based processes, migration from parceled and episode based healthcare provisioning to a workflow oriented model and increased awareness and responsibility of citizens towards their own health and wellness are key objectives of HeartDrive. In two scenarios for rehabilitation and home monitoring we show how the results are achieved by providing a solution that highlights a broader concept of cooperation that goes beyond technical interoperability towards semantic interoperability explicitly sharing process definitions, decision support strategies and information semantics.
Diagnosis is one of the most important processes in the medical field. Since the knowledge domain... more Diagnosis is one of the most important processes in the medical field. Since the knowledge domains of clinical specialties are expanding rapidly in terms of complexity and volume of data, clinicians have, in many cases, difficulties to make an accurate diagnosis. Therefore, intelligent and quantitative support for diagnostic tasks can be effectively exploited for improving the effectiveness of the process and reduce misdiagnosis. In this respect, Multi-Classifier Systems represent one of the most promising approaches within Machine Learning methodologies. This paper proposes a Multi-Classifier Systems framework for supporting diagnostic activities with the aim of improving diagnostic accuracy. The framework uses and combines several classification algorithms by dynamically selecting the most competent classifier according to the test sample and its location in the feature space. Here, we extend our previous research. The new experimental results, compared with several multi classifi...
Appointment scheduling systems represent a method to manage patient waiting lists effectively. Th... more Appointment scheduling systems represent a method to manage patient waiting lists effectively. This work advances an innovative quantitative approach for the outpatient appointment scheduling problems, based on an optimization model, to manage outpatient Day Service operations. It focuses on outpatient appointment scheduling. We start from earlier works in the literature to design models with the objective to maximize the number of patients' appointments, to reduce patient's waiting time, and to increase patient's satisfaction. The proposed combinatorial problem is solved by Answer Set Programming, which is a declarative logic formalism, widely used in Artificial Intelligence and recognized as a powerful tool for Knowledge Representation and Reasoning, to show the advantages of declarative programming for modelling and fast prototyping problem requirements. We apply the model to solve real-life scenarios of the Rheumatology domain. We compare the results on the real instance already solved in our earlier work and extend the computational experiments on some new generated and realistic instances. Since the computational times increase with the size of instances, we develop a three-phase solution approach based on patient's priority. The heuristic approach is hierarchical and enables to solve more instances than the one-run approach within the computational time limit.
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service... more This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. Highlights • We present integer programming models for the patient-bed assignment problem • We develop an efficient matheuristic to solve the NP-hard patient-bed assignment problem • We discuss on penalty values for soft models and improves all the best-known values in the literature
Background: To evaluate the utility of the revised coma remission scale (CRS-r), together with ot... more Background: To evaluate the utility of the revised coma remission scale (CRS-r), together with other clinical variables, in predicting emergence from a disorder of consciousness (DoC) after intensive rehabilitation care. Method: This is a prospective observational cohort study of consecutive 180 brain-injured patients with prolonged DoC upon admission to neurorehabilitation unit. 123 patients in a vegetative state (VS) and 57 in a minimally conscious state (MCS) were included and followed for a period of 8 weeks in the intensive care unit (ICU). Demographical and clinical factors were used as outcome measures. Univariate and multivariate Cox regression models were employed for examining potential predictors for clinical outcome along the time. Results: VS and MCS groups were matched for demographical and clinical (i.e., aetiology, tracheostomy and feed administration) variables. Within 2 months after admission in intensive neurorehabilitation unit, 3.9% were dead, 35.5% had a full r...
The application of artificial intelligence to extract predictors of Gambling disorder (GD) is a n... more The application of artificial intelligence to extract predictors of Gambling disorder (GD) is a new field of study. A plethora of studies have suggested that maladaptive personality dispositions may serve as risk factors for GD. Here, we used Classification and Regression Trees algorithm to identify multivariate predictive patterns of personality profiles that could identify GD patients from healthy controls at an individual level. Forty psychiatric patients, recruited from specialized gambling clinics, without any additional comorbidity and 160 matched healthy controls completed the Five-Factor model of personality as measured by the NEO-PI-R, which were used to build the classification model. Classification algorithm was able to discriminate individuals with GD from controls with an AUC of 77.3% (95% CI 0.65-0.88, p<0.0001). A multidimensional construct of traits including sub-facets of openness, neuroticism and conscientiousness was employed by algorithm for classification det...
2016 Computing in Cardiology Conference (CinC), 2016
Cardiac rehabilitation is a well-recognised nonpharmacological intervention that prevents the rec... more Cardiac rehabilitation is a well-recognised nonpharmacological intervention that prevents the recurrence of cardiovascular events. Previous studies investigated the application of data mining techniques for the prediction of the rehabilitation outcome in terms of physical, but fewer reports are focused on using predictive models to support clinicians in the choice of a patient-specific rehabilitative treatment path. Aim of the work was to derive a prediction model for help clinicians in the prescription of the rehabilitation program. We enrolled 129 patients admitted for cardiac rehabilitation after a major cardiovascular event. Data on anthropometric measures, surgical procedure and complications, comorbidities and physical performance scales were collected at admission. The prediction outcome was the rehabilitation program divided in four different paths. Different algorithms were tested to find the best predictive model. Models performance were measured by prediction accuracy. Mean model accuracy was 0.790 (SD 0.118). Best model selected was Lasso regression showing an average classification accuracy on test set of 0.935. Data mining techniques have shown to be a reliable tool for support clinicians in the decision of cardiac rehabilitation treatment path.
Good clinical governance of headache implies efficient and accessible diagnostic and therapeutic ... more Good clinical governance of headache implies efficient and accessible diagnostic and therapeutic paths involving health care at different levels [1]. Often clinicians do not appropriately assess and treat headache. Information and communication technologies might play a key role in improving access, quality, efficiency and prevention in health care. HealthSOAF (Service-Oriented Architecture Framework) is a networking and interoperability technological platform aimed to assist multiple level health care access and decision making. Its first real testing scenario in Europe has been the Headache Network in the Italian Region of Calabria targeting to assist clinicians at different levels of health care to correctly diagnose, manage and refer headache patients (Figure 1).
This paper presents an integrated location-allocation model balancing efficiency and equity crite... more This paper presents an integrated location-allocation model balancing efficiency and equity criteria. The new formulation combines two domains: facility location and data envelopment analysis. To support the decision maker with more realistic solutions based on the optimal location-allocation decisions, we endogenize the outputs of the model as a function dependent on the allocation variables. To illustrate the viability of the proposed approach, we investigated the potential application of the model to the design of an emergency medical service system.
Studies in health technology and informatics, 2006
HEARTFAID is a research and development project aimed at devising, developing and validating an i... more HEARTFAID is a research and development project aimed at devising, developing and validating an innovative knowledge based platform of services, able to improve early diagnosis and to make more effective the medical-clinical management of heart diseases within elderly population. Chronic Heart Failure is one of the most remarkable health problems for prevalence and morbidity, especially in the developed western countries, with a strong impact in terms of social and economic effects. All these aspects are typically emphasized within the elderly population, with very frequent hospital admissions and a significant increase of medical costs. Recent studies and experiences have demonstrated that accurate heart failure management programs, based on a suitable integration of inpatient and outpatient clinical procedures, might prevent and reduce hospital admissions, improving clinical status and reducing costs. HEARTFAID aims at defining efficient and effective health care delivery organiza...
In machine learning, hyperparameter tuning is strongly useful to improve model performance. In ou... more In machine learning, hyperparameter tuning is strongly useful to improve model performance. In our research, we concentrate our attention on classifying imbalanced data by cost-sensitive support vector machines. We propose a multi-objective approach that optimizes model’s hyper-parameters. The approach is devised for imbalanced data. Three SVM model’s performance measures are optimized. We present the algorithm in a basic version based on genetic algorithms, and as an improved version based on genetic algorithms combined with decision trees. We tested the basic and the improved approach on benchmark datasets either as serial and parallel version. The improved version strongly reduces the computational time needed for finding optimized hyper-parameters. The results empirically show that suitable evaluation measures should be used in assessing the classification performance of classification models with imbalanced data.
Springer Proceedings in Mathematics & Statistics, 2017
This work presents an innovative quantitative approach, based on an optimization model, to manage... more This work presents an innovative quantitative approach, based on an optimization model, to manage outpatient Day Service operations. The main objective of the research work is to maximize admitted patient flow in order to use available resources efficiently. The model is designed with the help of Answer Set Programming and implemented in Datalog with Disjunction (DLV) system. As significant case study, we consider the Rheumatology domain. Preliminary results are presented.
We propose a new set of clinical variables for a more accurate early prediction of safe decannula... more We propose a new set of clinical variables for a more accurate early prediction of safe decannulation in patients with severe acquired brain injury (ABI), during a post-acute rehabilitation course. Starting from the already validated DecaPreT scale, we tested the accuracy of new logistic regression models where the coefficients of the original predictors were reestimated. Patients with tracheostomy were retrospectively selected from the database of the neurorehabilitation unit at the S. Anna Institute of Crotone, Italy. New potential predictors of decannulation were screened from variables collected on admission during clinical examination, including (a) age at injury, (b) coma recovery scale-revised (CRS-r) scores, and c) length of ICU period. Of 273 patients with ABI (mean age 53.01 years; 34% female; median DecaPreT = 0.61), 61.5% were safely decannulated before discharge. In the validation phase, the linear logistic prediction model, created with the new multivariable predictors...
2018 Computing in Cardiology Conference (CinC), 2018
This work aims at developing and assessing a machine learning based Knowledge Discovery task for ... more This work aims at developing and assessing a machine learning based Knowledge Discovery task for risk prediction of major cardiovascular worsening events in chronic heart failure patients. Clinical data from 50patients with chronic heartfailure was analyzed. For each patient, personal data, different vital and clinical parameters and the presence of cardiovascular worsening events have been stored every three months per two years. We defined the Knowledge Discovery analysis as a predictive task stated as supervised binary classification problem. The class label was defined based on the occurrence or not of cardiovascular worsening events between two consecutive visits. To take into account the temporality of the worsening events, six different temporal weighting strategies, applied to the vital parameters, were tested. Several machine learning algorithms were applied for each strategy obtaining different predictive models. Models performance have been evaluated mainly in term of are...
Patient admission and surgery scheduling is a complex combinatorial optimization problem. It cons... more Patient admission and surgery scheduling is a complex combinatorial optimization problem. It consists on defining patient admission dates, assigning them to suitable rooms, and schedule surgeries accordingly to an existing master surgical schedule. This problem belongs to the class of NP-hard problems. In this paper, we firstly formulate an integer programming model for offline patient admissions, room assignments, and surgery scheduling; then apply a matheuristic that combines exact methods with rescheduling approaches. The matheuristic is evaluated using benchmark datasets. The experimental results improve those reported in the literature and show that the proposed method outperforms existing techniques of the state-of-the-arts.
This paper originates from the HeartDrive project, a platform of services for a more effective, e... more This paper originates from the HeartDrive project, a platform of services for a more effective, efficient and integrated management of heart failure and comorbidities. HeartDrive establishes a cooperative approach based on the concepts of continuity of care and extreme, patient oriented, customization of diagnostic, therapeutic and follow-up procedures. Definition and development of evidence based processes, migration from parceled and episode based healthcare provisioning to a workflow oriented model and increased awareness and responsibility of citizens towards their own health and wellness are key objectives of HeartDrive. In two scenarios for rehabilitation and home monitoring we show how the results are achieved by providing a solution that highlights a broader concept of cooperation that goes beyond technical interoperability towards semantic interoperability explicitly sharing process definitions, decision support strategies and information semantics.
Diagnosis is one of the most important processes in the medical field. Since the knowledge domain... more Diagnosis is one of the most important processes in the medical field. Since the knowledge domains of clinical specialties are expanding rapidly in terms of complexity and volume of data, clinicians have, in many cases, difficulties to make an accurate diagnosis. Therefore, intelligent and quantitative support for diagnostic tasks can be effectively exploited for improving the effectiveness of the process and reduce misdiagnosis. In this respect, Multi-Classifier Systems represent one of the most promising approaches within Machine Learning methodologies. This paper proposes a Multi-Classifier Systems framework for supporting diagnostic activities with the aim of improving diagnostic accuracy. The framework uses and combines several classification algorithms by dynamically selecting the most competent classifier according to the test sample and its location in the feature space. Here, we extend our previous research. The new experimental results, compared with several multi classifi...
Appointment scheduling systems represent a method to manage patient waiting lists effectively. Th... more Appointment scheduling systems represent a method to manage patient waiting lists effectively. This work advances an innovative quantitative approach for the outpatient appointment scheduling problems, based on an optimization model, to manage outpatient Day Service operations. It focuses on outpatient appointment scheduling. We start from earlier works in the literature to design models with the objective to maximize the number of patients' appointments, to reduce patient's waiting time, and to increase patient's satisfaction. The proposed combinatorial problem is solved by Answer Set Programming, which is a declarative logic formalism, widely used in Artificial Intelligence and recognized as a powerful tool for Knowledge Representation and Reasoning, to show the advantages of declarative programming for modelling and fast prototyping problem requirements. We apply the model to solve real-life scenarios of the Rheumatology domain. We compare the results on the real instance already solved in our earlier work and extend the computational experiments on some new generated and realistic instances. Since the computational times increase with the size of instances, we develop a three-phase solution approach based on patient's priority. The heuristic approach is hierarchical and enables to solve more instances than the one-run approach within the computational time limit.
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service... more This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. Highlights • We present integer programming models for the patient-bed assignment problem • We develop an efficient matheuristic to solve the NP-hard patient-bed assignment problem • We discuss on penalty values for soft models and improves all the best-known values in the literature
Background: To evaluate the utility of the revised coma remission scale (CRS-r), together with ot... more Background: To evaluate the utility of the revised coma remission scale (CRS-r), together with other clinical variables, in predicting emergence from a disorder of consciousness (DoC) after intensive rehabilitation care. Method: This is a prospective observational cohort study of consecutive 180 brain-injured patients with prolonged DoC upon admission to neurorehabilitation unit. 123 patients in a vegetative state (VS) and 57 in a minimally conscious state (MCS) were included and followed for a period of 8 weeks in the intensive care unit (ICU). Demographical and clinical factors were used as outcome measures. Univariate and multivariate Cox regression models were employed for examining potential predictors for clinical outcome along the time. Results: VS and MCS groups were matched for demographical and clinical (i.e., aetiology, tracheostomy and feed administration) variables. Within 2 months after admission in intensive neurorehabilitation unit, 3.9% were dead, 35.5% had a full r...
The application of artificial intelligence to extract predictors of Gambling disorder (GD) is a n... more The application of artificial intelligence to extract predictors of Gambling disorder (GD) is a new field of study. A plethora of studies have suggested that maladaptive personality dispositions may serve as risk factors for GD. Here, we used Classification and Regression Trees algorithm to identify multivariate predictive patterns of personality profiles that could identify GD patients from healthy controls at an individual level. Forty psychiatric patients, recruited from specialized gambling clinics, without any additional comorbidity and 160 matched healthy controls completed the Five-Factor model of personality as measured by the NEO-PI-R, which were used to build the classification model. Classification algorithm was able to discriminate individuals with GD from controls with an AUC of 77.3% (95% CI 0.65-0.88, p<0.0001). A multidimensional construct of traits including sub-facets of openness, neuroticism and conscientiousness was employed by algorithm for classification det...
2016 Computing in Cardiology Conference (CinC), 2016
Cardiac rehabilitation is a well-recognised nonpharmacological intervention that prevents the rec... more Cardiac rehabilitation is a well-recognised nonpharmacological intervention that prevents the recurrence of cardiovascular events. Previous studies investigated the application of data mining techniques for the prediction of the rehabilitation outcome in terms of physical, but fewer reports are focused on using predictive models to support clinicians in the choice of a patient-specific rehabilitative treatment path. Aim of the work was to derive a prediction model for help clinicians in the prescription of the rehabilitation program. We enrolled 129 patients admitted for cardiac rehabilitation after a major cardiovascular event. Data on anthropometric measures, surgical procedure and complications, comorbidities and physical performance scales were collected at admission. The prediction outcome was the rehabilitation program divided in four different paths. Different algorithms were tested to find the best predictive model. Models performance were measured by prediction accuracy. Mean model accuracy was 0.790 (SD 0.118). Best model selected was Lasso regression showing an average classification accuracy on test set of 0.935. Data mining techniques have shown to be a reliable tool for support clinicians in the decision of cardiac rehabilitation treatment path.
Good clinical governance of headache implies efficient and accessible diagnostic and therapeutic ... more Good clinical governance of headache implies efficient and accessible diagnostic and therapeutic paths involving health care at different levels [1]. Often clinicians do not appropriately assess and treat headache. Information and communication technologies might play a key role in improving access, quality, efficiency and prevention in health care. HealthSOAF (Service-Oriented Architecture Framework) is a networking and interoperability technological platform aimed to assist multiple level health care access and decision making. Its first real testing scenario in Europe has been the Headache Network in the Italian Region of Calabria targeting to assist clinicians at different levels of health care to correctly diagnose, manage and refer headache patients (Figure 1).
This paper presents an integrated location-allocation model balancing efficiency and equity crite... more This paper presents an integrated location-allocation model balancing efficiency and equity criteria. The new formulation combines two domains: facility location and data envelopment analysis. To support the decision maker with more realistic solutions based on the optimal location-allocation decisions, we endogenize the outputs of the model as a function dependent on the allocation variables. To illustrate the viability of the proposed approach, we investigated the potential application of the model to the design of an emergency medical service system.
Studies in health technology and informatics, 2006
HEARTFAID is a research and development project aimed at devising, developing and validating an i... more HEARTFAID is a research and development project aimed at devising, developing and validating an innovative knowledge based platform of services, able to improve early diagnosis and to make more effective the medical-clinical management of heart diseases within elderly population. Chronic Heart Failure is one of the most remarkable health problems for prevalence and morbidity, especially in the developed western countries, with a strong impact in terms of social and economic effects. All these aspects are typically emphasized within the elderly population, with very frequent hospital admissions and a significant increase of medical costs. Recent studies and experiences have demonstrated that accurate heart failure management programs, based on a suitable integration of inpatient and outpatient clinical procedures, might prevent and reduce hospital admissions, improving clinical status and reducing costs. HEARTFAID aims at defining efficient and effective health care delivery organiza...
Uploads
Papers by Domenico Conforti