PhD in DATA ANALYTICS AND DECISION SCIENCES - 36th cycle · educated to become a skilful health analyst, i.e., a data scientist with deep expertise in statistical modeling for the

PhD in DATA ANALYTICS AND DECISION SCIENCES -

36th cycle

Number of scholarship offered 6

DepartmentDIPARTIMENTO DI ELETTRONICA, INFORMAZIONE EBIOINGEGNERIA

Description of the PhD Programme

The PhD program in Data Analytics and Decision Sciences (DADS) aims at training highlyqualified senior data analysts and data managers capable of carrying out research at universities,international institutions, tech and financial companies, regulatory authorities, and other publicbodies. The program stems from the cooperation between three departments: Dipartimento diElettronica, Informazione e Bioingegneria (DEIB), Dipartimento di Ingegneria Gestionale (DIG),Dipartimento di Matematica (DMAT), and the Center for Analysis, Decisions and Society (CADS)at Human Technopole. It gives the enrolled students the opportunity to work in a highlyinterdisciplinary environment with strong connections to international research centers and privatecompanies. The program provides successful candidates with the opportunity to acquire a highdegree of professional expertise in specific scientific and technological fields. The program laststhree years: upon its successful completion and final exam, candidates will be awarded the title ofPhD in Data Analytics and Decision Sciences. The first year is devoted to the courses that buildthe broad competence and the strong interdisciplinary set of skills required by data analytics. Thenext two years focus on the development of the Doctoral thesis. Students are required to spend atleast one semester in a research institution abroad, taking advantage of the network ofinternational collaborations of the three departments involved in the program. All the studentsenrolled in the DADS Doctoral Program are supported by scholarships from public institutions andprivate companies.

Stampato il 09/04/2020 1 / 1


36th cycle

Research Field: DATA ANALYTICS AND DECISION SCIENCES

Monthly net income of PhDscholarship (max 36 months)

€ 1300.0In case of a change of the welfare rates during the three-year period, the amount could be modified.

Context of the research activity

Motivation and objectives of the researchin this field

The PhD program in Data Analytics and DecisionSciences aims at breeding the next generation of datascientists who will tackle the challenges and theopportunities created by the increasingly larger availabilityof data, cheap storage, and computer power. Datascientists who can capture the most relevant aspects ofphenomena in play, develop adequate models, supervisethe development of analytic pipelines, critically analyzethe result, and support the technological transfer that theresults might enable.

Methods and techniques that will bedeveloped and used to carry out theresearch

The research aims at developing novel approaches fordata analytics and decision science including novelstatistical methods for complex data, new econometricsetups, methods of advanced machine learning, as wellas technological solutions to integrate massive amount ofdata from heterogeneous data sources.

Educational objectives

The research will be carried out in collaboration with theteam of applied statisticians, economists, computer andmanagement engineers from three departments, theDepartment of Management, Economics and IndustrialEngineering, the Department of Mathematics, and theDeparment of Electronics, Information andBioengineering. Within this lively and stimulatingacademic and research environment, the doctoral student

1 / 2

will be educated to become a skilful researcher in DataScience.

Job opportunities

The profile of Data Scientist and the technological andmethodological solutions proposed are of interest for abroad range of actors, including International institutions,policy makers, statistical and research centres, largecorporations that aim to analyse heterogeneous datacoming from various datasources.

Composition of the research group

9 Full Professors8 Associated Professors0 Assistant Professors11 PhD Students

Name of the research directors Any faculty member

Contacts

PhD CoordinatorProf. Pier Luca LanziEmail: [email protected]: +390223993472

Additional support - Financial aid per PhD student per year (gross amount)

Housing - Foreign Students --

Housing - Out-of-town residents(more than 80Km out of Milano) --

Additional information: educational activity, teaching assistantship, computer availability, desk availability,any other information

There are various forms of financial aid for activities of support to the teaching practice. The PhDstudent is encouraged to take part in these activities, within the limits allowed by the regulations.

Increase in the scholarship for stays abroad: Euro 566,36 per month, for up to 6 months

2 / 2


36th cycle

Research Field: DEEP LEARNING SEGMENTATION AND FEATURE EXTRACTION FOR

MEDICAL IMAGING





Medical imaging plays a key role in the diagnostic processas well as in the choice of the most appropriatetreatments for each patient. Unfortunately, the analysis,the segmentation, and the extraction of features frommedical images is still a very time-consuming process.Deep Learning is a promising approach to automate andstandardize these tasks.This research aims at the designand application of novel deep learning methods torelevant medical imaging processing tasks, such as thesegmentation of organs at risk for radiotherapy planning,the tumor segmentation in MRI/PET/CAT images, and theextraction of predictive features from segmented lesions.


To achieve the goals of this research project, novelmethods will be devised on the basis of the latest findingsin Deep Learning. In particular, to deal with the limitedamount of annotated data, methods such as GenerativeAdversarial Networks and Cycle-Consistent AdversarialNetworks will be used for data augmentation and fortraining the segmentation models in an unsupervisedlearning setting. Approaches based on VariationalAutoencoder will be employed to extract predictivefeatures from the segmented images. In addition,an Active Learning framework will be used to reduce thehuman effort in the data labelling process and to validatethe outputs of the segmentation models designed.

1 / 3


The PhD is expected to develop multidisciplinaryexpertise in the fields of Deep Learning and Health DataScience, as this research will be supervised by a machinelearning expert at Politecnico di Milano. The research willalso involve clinicians and researchers of HumanitasUniversity and CADS - Center for Analysis Decisions andSociety, the joint research center of Human Technopoleand Politecnico di Milano. Within this lively and stimulatingacademic and research environment, the student will beeducated to become a skilful health analyst, i.e., a datascientist with deep expertise in statistical modeling for thehealthcare setting.

Job opportunities

The expertise and skills developed in this PhD open toseveral opportunities at national and international level, inacademic and industrial research as well as in companies,in e-health and in any domain where data science andmachine learning is relevant.



Name of the research directors Daniele Loiacono, Arturo Chiti

Contacts

Daniele Loiacono, [email protected],+390223993615, http://home.deib.polimi.it/loiacono/

Arturo Chiti, [email protected],+3902 82241 6621,https://www.hunimed.eu/member/arturo-chiti/

List of Universities, Companies, Agencies and/or National or International Institutions that arecooperating in the research

Human Technopole•

Humanitas University•

2 / 3







3 / 3


36th cycle

Research Field: IMPACT ANALYSIS IN REAL WORLD CLINICAL ONCOLOGY





The research stream addresses the challengesassociated with causal inferences and impact analysis inhealth care settings. We envisage to integrate genomic,clinical and economic data and to perform healthtechnology assessment analyses based on thecombination between predictive statistical methods andcausal models. Research will be performed incollaboration with a research team at the Istituto Europeodi Oncologia, IEO. Specifically, we aim to develop anoriginal framework for cost benefit analysis in ovarian andlung cancers, i.e. two disease areas of paramountimportance.


The impact of molecular diagnostic tools will bemeasured, so to analyze the sensitivity of cost benefitassessments for substitute, or complementary,treatments, while relaxing assumptions of homogeneity forthe underlying patient population. The research will focuson the development of statistical and predictive models,machine learning algorithms, to estimate social costs onthe basis, for example, of the patient's follow- up (estimateof indirect costs related to the patient and caregiver forexample), taking into account both indirect costs andavoided losses of working time and productivity.

Educational objectives The research will be carried out in collaboration with the

1 / 3

team of clinicians at IEO and, moreover, with the team ofapplied statisticians, economists, computer andmanagement engineers who are collaborating with theCenter for Analysis Decision and Society of the HumanTechnopole. The student will be educated to become askillful health data scientist, with statistical, econometricand economic competences

Job opportunities

The profile of data scientist/econometrician and theactivities proposed in this project are of interest to a broadrange of actors, including (but not limited to): public andprivate institutions dealing with clinical andpharmaceutical data, as well as international institutionsand research centers aiming at assessing outcomes andeconomic impact in health.



Name of the research directors Prof. Anna Maria Paganoni, Prof. Fabio Pammolli

Contacts

Prof. Anna Maria Paganoni, [email protected],0223994574, https://www.mate.polimi.it/?view=pp&id=65&lg=it#ann Prof. Fabio Pammolli, [email protected]. Pier Giuseppe Pelicci, IEODott. Luca Mazzarella, IEO


Human Technopole (HT)•

CHRP - Center for Healthcare Research and Pharmacoepidemiology•

IEO - Istituto Europeo di Oncologia•

SOSE - Soluzioni per il Sistema Economico pubblico e privato•


2 / 3






3 / 3


36th cycle

Research Field: LEARNING ANALYTICS TO SUPPORT POLICIES AND MANAGERIAL

DECISION MAKING IN EDUCATIONAL INSTITUTIONS





The research, along the three years, focuses on theanalysis and prediction of performance of students fromsecondary schools and universities. Moreover, theeffectiveness of remedial education interventions aretested to assess the causality nexus betweenpolicies/interventions with students' careers. Every activityis preceded by a deep study of the existing literatureabout the topic and through the development of a referredtheoretical framework, driving the analysis. Specifically,the research is realized along three main lines.

The first research stream deals with the prediction of at-risk students from various academic institutions (includingPolitecnico di Milano), through the analysis ofadministrative data. In this phase, Machine Learningmodels aretested together with more traditional (statisticaland econometric) ones, to assure high-predictionaccuracy performance. This approach allowsimplementing such algorithms making real-time forecasts.In the models' development phase, it is important toconsider the timing of prediction, because this can identifypossible withdrawing students as soon as possible, givinginstitutions time to set retention strategies. The secondresearch stream copes with predictions of academicperformance for secondary school students.

In this case, too, the development of "Early Warning1 / 4

Systems" is extremely important in giving schools time tointervene and test the efficacy of such activities. In thecase of secondary schools, specific models are developedto take into account the availability of weekly data.Thethird research stream is devoted to the assessment ofeducational interventions to retain and support students.Indeed, the main advantage given by prediction modellingfor schools and universities is the possibility to setinterventions to improve at-risk students' careers. Theevaluation of these interventions is possible throughexperimenting different strategies and evaluating thecausality nexus between the intervention and thestudent's academic results.


The research uses a variety of econometric, statistical andmachine learning approaches - in many circumstances,cross-validating the results and mixing the methodsthemselves. Among others we cite:

Econometric methods•

Randomized experiments•

Propensity score matching•

Fixed Effects regressions•

Instrumental Variables•

Regression DiscontinuityStatistical methods•

Generalized linear models•

Multilevel models•

Latent Class ModelsMachine Learning techniques•

Random trees•

Random Forests•

CART•

Neural networks•


The research is carried out in strict collaboration withHigher Education decision-makers, school principals andeducators. The student will be educated to become askillful educational data scientist, with statistical,econometric and economic competences.

2 / 4

Job opportunities

The profile of data scientist/econometrician and theactivities proposed in this project are of interest to a broadrange of actors, including (but not limited to): public andprivate institutions dealing with educational institutions, aswell as international institutions and research centersaiming at assessing outcomes and economic impact ineducation.



Name of the research directors Prof. Anna Maria Paganoni, Prof. Tommaso Agasisti

Contacts

Prof. Anna Maria Paganoni, [email protected],0223994574, https://www.mate.polimi.it/?view=pp&id=65&lg=it#ann

Prof. Tommaso Agasisti, [email protected], 0223993963


INVALSI - Istituto nazionale per la valutazione del sistema educativo di istruzione e di

formazione

•

Amsterdam Center for Learning Analytics (ACLA)•

Teachers College - Columbia University4. Fondazione per la Scuola•






3 / 4


4 / 4


36th cycle

Research Field: NEUROSCIENCES AND GENOMICS: ADVANCED DATA ANALYTICS

METHODS





Neurosciences offer highly complex data coming fromneuroimaging modalities such as functional MagneticResonance Imaging (fMRI) andMagnetoEncephaloGraphy (MEG). The analysis of thesecomplex data poses new challenging theoretical andmethodological problems to data science, and calls for thedefinition of new data analytics methods, mergingapproaches and ideas from statistics and from computersciences. Moreover, neuroimaging data are oftenenriched by genomics information. Data integration fromthese heterogenous sources may lead to groundbreakingadvances in our knowledge of neuronal disorders,supporting precision medicine in this context.


The PhD candidate will define new functional dataanalysis methods for complex data arising from currentneuroimaging modalities such as fMRI and MEG. Thesewill include regression and classification methods forthese high-dimensional functional data, and will alsoenable the integration of genomics information. Themethods will be computationally highly efficient,leveraging on heterogenous system approaches.

Educational objectives The research will be carried out in collaboration with theteam of statisticians, computer scientists, economists and

1 / 3

management engineers cooperating with the Center forAnalysis Decision and Society of the Human Technopole.Within this lively and stimulating academic and researchenvironment, the doctoral student will become a skilfuldata scientist, with expertise in both advanced statisticallearning and advanced computer science methods.

Job opportunities

Data scientists and data analytics experts currently arethe most in-demand job today, among high- qualificationjobs. In all industrial and business sectors, the demand fordata scientists and analytics experts continues to outpacesupply and dominates both the US and the European jobmarket.



Name of the research directors L.M. Sangalli, D. Sciuto, M.D. Santambrogio

Contacts

Prof. Laura M. Sangalli: [email protected], 02 23994554, http://mox.polimi.it/users/sangalliProf. Donatella Sciuto: [email protected], 02 23993662, http://home.deib.polimi.it/sciuto/ Prof. Marco D. Santambrogio: [email protected] 0223994012 http://home.deib.polimi.it/santambr/

List of Universities, Companies, Agencies and/or National or International Institutions that arecooperating in the research:

Human Technopole•





There are various forms of financial aid for activities of support to the teaching practice. The PhD2 / 3

student is encouraged to take part in these activities, within the limits allowed by the regulations.


3 / 3


36th cycle

Research Field: PATIENT REPRESENTATION METHODS FOR RISK STRATIFICATION IN

ONCOLOGY





A primary goal of precision medicine is to developquantitative models for patients that can be used topredict health status, as well as to help prevent disease ordisability. In this context, Oncology is a complex medicalfield where combining treatment history, genetic markers,medical imaging and personal information couldsignificantly improve patients' response to cures.However, dealing with patients data (multimodal, smallsample size, high dimensionality and noise) calls for novelstatistical and machine learning approaches to be turnedinto actionable medical insights. Representation learningis a broad mathematical field dealing with learning theintrinsic structure of data so that it can be reshaped andpresented in a way that makes it easier to extractinsightful knowledge. Indeed, oncological patients' data inthe form of free text, images, administrative data andelectronic health records (HER) should be handledproperly to be combined and create the best 360 degreespatient's representation.


The research will focus on the design and development ofnovel statistical and representation learning methods todeal with multimodal data sources in the field ofoncological personalized medicine. It will covertechniques spanning from traditional statisticalmethodologies, functional data analysis, imageprocessing, to the novel and complex deep learning

1 / 3

techniques, and the researcher will apply the propermathematical tools to represent patients¿ data forexample to (i) predict response to cancer treatment; (ii)stratify patients; (iii) learn medical concepts from EHR orfree text, or (iv) identify patterns and/or specific featuresthat could provide actionable insights to clinicians.


The PhD is expected to develop multidisciplinaryexpertise in the fields of Statistical Learning and HealthData Science, as this research will be supervised by adata science expert at Politecnico di Milano. The researchwill also involve clinicians and researchers of IstitutoNazionale dei Tumori and CADS - Center for AnalysisDecisions and Society, the joint research center of HumanTechnopole and Politecnico di Milano. Within this livelyand stimulating academic and research environment, thestudent will be educated to become a skilful healthanalyst, i.e., a data scientist with deep expertise instatistical modeling for the healthcare setting.

Job opportunities

The profile of data scientist and the applications proposedin this project are of interest to of a broad range of actors,including (but not limited to): public and private institutionsdealing with healthcare, hospitals, clinical andpharmaceutical companies, as well as internationalinstitutions and research centres working in healthcareresearch, and policy makers in charge with healthcaregovernance.



Name of the research directors Prof. Francesca Ieva, Prof. Giovanni Corrao

Contacts

Prof. Francesca Ieva [email protected], 0223994578,https://sites.google.com/view/francesca-ieva/home


Human Technopole•2 / 3

Humanitas•

Center for Healthcare Research and Pharmacoepidemiology•







3 / 3


36th cycle

Research Field: STANDARDIZED-SPENDING AND CONVERGENCE POLICY IN

DECENTRALIZED HEALTHCARE SYSTEMS: ITALY





Our research aims at developing a theoretical andempirical support for policy design in a context ofdecentralized decision-making, at different levels(Regions, hospitals, etc.). It specifically focuses on 1) thedevelopment of a standardized-spending estimationmethodology, so as to support allocation of resources in adecentralized context consistently with relevant objectivesat the national level (efficiency, equity, effectiveness ofcare); 2) the analysis of the "local" departures from thestandardized benchmarks and of their relevantdeterminants, so as to identify effective policy actions forconvergence.


The research activity will explore the complex web ofissues underlying its two main objectives. It willspecifically focus on 1) the measurement of the outputand inputs of healthcare production through the use ofappropriate composite indicators, which are the basic datafor the implementation of the other research activities; 2)the estimation of a cost function, based on a previousestimation of a standard output/demand function, whichsupports the standardized-spending estimationmethodology; 3) the estimation of technical efficiency,through a combination of data envelopment analysis andeconometric analysis, and the measurement of an outputgap, through the comparison of actual with standard

1 / 3

output, which will be the components of the analysis of thedepartures from the standardized benchmarks.


The research will be carried out in collaboration withexperts at SOSE (Ministry of Economy and Treasury) and,moreover, with the team of applied statisticians,economists, computer and management engineerscollaborating with the Center for Analysis Decision andSociety of the Human Technopole. The student will beeducated to become a skillful public health economist.

Job opportunities

The profile of health economist and the researchproposed in this project are of interest to of a broad rangeof actors, including (but not limited to): public and privateinstitutions dealing with healthcare data, as well asinternational institutions and research centers aiming atassessing performances of healthcare providers, andpolicy makers in charge with healthcare expenditure andgovernance.



Name of the research directors Prof. Fabio Pammolli, Prof. Giacomo Pignataro

Contacts

Prof. Fabio Pammolli, [email protected]. Giacomo Pignataro, [email protected]. Francesco Porcelli, Università di BariDott. Francesco Vidoli, SOSE


Human Technopole (HT)•

Università di Bari•

Università di Catania•

SOSE - Soluzioni per il Sistema Economico pubblico e privato•

2 / 3







3 / 3

Documents

PhD in DATA ANALYTICS AND DECISION SCIENCES - 36th cycle · educated to become a skilful health analyst, i.e., a data scientist with deep expertise in statistical modeling for the