Search

Research in innovation studies usually relies on financial statements, surveys, or patents as primary data sources, although these sources of information show some limitations when applied to Small and Medium Enterprises (SMEs). Our paper explores whether the HTML code of a company's website is a further source to better inform innovation policies, under the assumption that how HTML is employed in crafting a corporate website provides insights into the company's innovation capabilities. In particular, we leverage HTML tags and their associations to empirically show that the websites of innovative SMEs are different from non-innovative ones both in terms of their size and coding practices. Our findings, based on a sample of Italian companies, indicate that the features of the HTML code of corporate websites reflect unobservable characteristics related to the skills and creativity present in businesses.

This paper addresses the importance of industry-specific models for SMEs bankruptcy prediction, building on earlier research finding larger predictive accuracy and enhanced temporal stability. Using Italian data, we propose separate bankruptcy prediction models for a few industries based on balance sheet data and explore the predictive power of SMEs' website html code structure. Our findings suggest that website data can serve as a valid complementary source for bankruptcy prediction, with different performances across sectors. We observe a certain degree of sectoral heterogeneity in the importance of financial ratios, firm-specific characteristics, and website structure, calling for an industry-tailored approach in bankruptcy prediction models.

Introduction: COVID-19 features changed with the Omicron variant of SARS-CoV-2 in adults. This study aims to describe COVID-19 symptoms in children and adolescents during the Parental, Delta, and Omicron eras Methods: A single-centre, prospective observational study was conducted on individuals aged 0–20 years attending the University Hospital of Padua (Italy) from April 2020 to December 2022. COVID-19 cases were defined by positive SARS-CoV-2 molecular detection and/or serology; patient/family symptoms and virological positivity were considered to determine the infection onset. Variables were summarized and compared using appropriate tests of descriptive statistics Results: A total of 509 cases [46% female, median age eight years (IQR: 4–12)] were studied. Three-hundred-eighty-seven (76%), 52 (10%), and 70 (14%) subjects experienced COVID-19 during the Parental, Delta, and Omicron waves, respectively. All subjects developed an asymptomatic/mild COVID-19. Overall, the most frequent symptoms were fever (47%) and rhinitis (21%), which showed a significant increasing incidence from the Parental to Omicron waves (p < 0.001). Conversely, diarrhea was most common during the pre-Omicron eras (p = 0.03). Stratifying symptoms according to the age group, fever, rhinitis, and skin rashes were observed more frequently among infants/toddlers; conversely, fatigue was more common in children older than five years. The duration of symptoms was similar across different SARS-CoV-2 variants of concern (VOCs); conversely, the number of symptoms varied according to the age group (p < 0.0001) Discussion: This study showed differences in COVID-19 clinical presentation among infants, children, and adolescents and confirmed Omicron infection is more likely to be associated with upper respiratory symptoms. However, further population-based studies are needed to support these findings. In addition, active surveillance will play a crucial role in assessing the disease severity of future

Aim: Children have largely been unaffected by severe COVID-19 compared to adults, but data suggest that they may have experienced new conditions after developing the disease. We compared outcomes in children who had experienced COVID-19 and healthy controls. Methods: A retrospective nested cohort study assessed the incidence rate of new-onset conditions after COVID-19 in children aged 0–14 years. Data were retrieved from an Italian paediatric primary care database linked to Veneto Region registries. Exposed children with a positive nasopharyngeal swab were matched 1:1 with unexposed children who had tested negative. Conditional Cox regression was fitted to estimate the adjusted hazard ratios (aHR) and 95% confidence intervals (CI) for the exposure and outcome associations after adjusting for covariates. Results: We compared 1656 exposed and 1656 unexposed children from 1 February 2020 to 30 November 2021. The overall excess risk for new-onset conditions after COVID-19 was 78% higher in the exposed than unexposed children. We found significantly higher risks for some new conditions in exposed children, including mental health issues (aHR 1.8, 95% CI 1.1–3.0) and neurological problems (aHR 2.4, 95% CI 1.4–4.1). Conclusion: Exposed children had a 78% higher risk of developing new conditions of interest after COVID-19 than unexposed children.

Academic research and the financial industry have recently shown great interest in Machine Learning algorithms capable of solving complex learning tasks, although in the field of firms' default prediction the lack of interpretability has prevented an extensive adoption of the black-box type of models. In order to overcome this drawback and maintain the high performances of black-boxes, this paper has chosen a model-agnostic approach. Accumulated Local Effects and Shapley values are used to shape the predictors' impact on the likelihood of default and rank them according to their contribution to the model outcome. Prediction is achieved by two Machine Learning algorithms (eXtreme Gradient Boosting and FeedForward Neural Networks) compared with three standard discriminant models. Results show that our analysis of the Italian Small and Medium Enterprises manufacturing industry benefits from the overall highest classification power by the eXtreme Gradient Boosting algorithm still maintaining a rich interpretation framework to support decisions.

Small and Medium Enterprises (SMEs) contribution to the European Union economy has always been relevant, for both value added and the creation of jobs. That is why the prediction of their survival is considered one of the economic pillars UE keeps under observation. Default prediction models, accounting for SMEs idiosyncratic traits, are based on several types of data, mainly accounting indicators. Balance sheet data, indeed, are considered the standard predictors for classification models in this field, although they do not allow to completely overcome the information opacity that is one of the main barriers preventing these firms from accessing credit. In our work, we explore the possibility of complementing accounting information with data scraped from the firms’ websites. We modeled the data using a nonlinear discriminant analysis and we benchmarked the results with the Logistic Regression. The evidence of our study is promising although the combination of online and offline data shows better results in case of survival firms than for defaulted companies.

SARS-CoV-2 infection may impair behavior and mental health; we explored the psychological impact on parents and children who experienced COVID-19 within their families. A cross-sectional web-based survey was conducted on families attending the COVID-19 Follow-up Clinic at the Department for Women’s and Children’s Health, Padua (Italy). From April 2020 to August 2021, 75 surveys were collected from 66 families (97 parents and 129 children); almost 70% of participants had COVID-19, mostly asymptomatic/mildly symptomatic, and the median time from infection to survey compilation was 164.7 days (SD 56). Most parents (>87%) reported positive relationships with family members either before, during, or after COVID-19. More than one-third of children and adolescents were unable to adapt to isolation. Among 31 pre-school children with a median age of 3 (SD 1.7), a change of one or more functions was reported for 74.2% of cases irrespective of COVID-19 status, particularly a change in circadian rhythm (25%), in relationship with parents (42.8%), and poor emotional control (36%). Among 74 children with a median age of 10.9 years (SD 2.7), 8.1% had a score indicating a disease; however, significant impairment in attention was reported for 16.7%, along with anxiety/depression and problems with conduct in 5.6% and 6.5% of cases, respectively.

Small and medium enterprises (SMEs) contribution to the European Union economy has always been relevant, for both value added and the creation of jobs. On the same time, SMEs are more fragile and likely to default as compared to larger firms so to induce public policies targeted on them. Specific default predic- tion models, accounting for SMEs idyiosincratic traits, are based on several types of data, mainly accounting indicators. We explore the possibility of complementing accounting information with data scraped from the firms’ websites.

Research in innovation usually builds on conventional data such as balance sheets, surveys, patents, or product catalogs. This paper intends to explore unconventional data, specifically web-scraped data, as an information source for innovation studies, proposing a careful procedure to establish the veracity of the linkage between web-based data and firm-level information retrieved from conventional sources. The study regards a sample of Italian manufacturing small and medium enterprises active in 2016, comprehending both innovative and non-innovative firms. It is based on HTML tags, whilst most of the previous literature worked on the web-pages text and related semantics. Our paper provides evidence that the way HTML language is applied to build a corporate website unveils the capabilities of the owner firm, helping to distinguish innovative from non-innovative SMEs.

L’obiettivo di un modello di rischio di credito e` quello di sviluppare una regola accurata in grado di distinguere tra osservazioni buone e cattive. In questo lavoro le aziende sono le nostre osservazioni, in particolare ci concentriamo sulle Piccole e Medie Imprese (PMI), come la maggior parte dei lavori nella letteratura finanziaria, dato il loro contributo all’economia dell’Unione Europea sia per il val- ore aggiunto sia per la creazione di posti di lavoro. Lo scopo dell’articolo e` duplice: esplorare l’uso degli indicatori ottenuti dal web per migliorare le previsioni di in- solvenza delle imprese e confrontare le prestazioni dei modelli alimentati con tali indicatori rispetto a quelli costruiti con le informazioni finanziarie standard. I risul- tati, ottenuti su un campione di aziende spagnole, includeranno anche un focus sulla capacita` degli indicatori web di riconoscere le aziende insolventi non comuni., The objective of a Credit Risk model is to develop an accurate rule that can distinguish between good and bad instances. In this work instances are firms, specif- ically we focus on Small and Medium Enterprises (SMEs), as most of the papers in the financial literature, thanks to their contribution to the European Union economy for both value added and the creation of jobs. The aim of the paper is twofold: to explore the usage of the web-scraped indicators to improve firms’ default predic- tions and to compare the models’ performances fed with the web-based indicators with the ones built with the standard financial information. The results, obtained on a sample of Spanish companies, will also include a focus about the ability of the web features in recognizing uncommon defaulted firms.

Recent literature shows an increasing interest in considering alternative sources of information for predicting Small and Medium Enterprises default. The usage of accounting indicators does not allow to completely overcome the information opacity that is one of the main barriers preventing these firms from accessing to credit. This complicates matters both for private lenders and for public institutions supporting policies. In this paper we propose corporate websites as an additional source of information, ready to be exploited in real-time. We also explore the joint use of online and offline data for enhancing correct prediction of default through a Kernel Discriminant Analysis, keeping the Logistic Regression and the Random Forests as benchmark. The obtained results shed light on the potentiality of these new data when accounting indicators lead to a wrong prediction.

Tail events, due to their nature, occur very rarely and are therefore often regarded as an element of surprise. However, business continuity management (BCM) as a part of enterprise risk management (ERM) should protect businesses from disruption in case of such events. In this chapter, we investigate whether businesses of different sizes were ready to withstand the consequences of the outbreak of the global SARS-CoV-2 pandemic in 2020. To this aim, we analyse the results of a global survey conducted during a major lockdown period (April-May 2020) by applying multiple correspondence analysis (MCA) and clustering. Our findings indicate that the financial sector was the best prepared for tail events, which may be attributed to regulations enforced for most financial institutions. Although around a half of businesses did not have a business continuity plan (BCP), having one helped adjust to the new conditions efficiently.

The paper explores the possibility to employ the source code of corporate websites as an information source for research in innovation studies. Research in this area is generally based on studies that collect data on patents or official data sources. Our paper links the standard economic information of the firm with web-based data and joins the ongoing debate with a threefold contribution. First, whereas the majority of the literature focused on the linguistic content of web-pages, we mostly use HTML tags. Second, we propose a method to assess the quality of the linkage of Web data to firm-level information. Third, we show that the data retrieved from corporate websites can aid to identify 'innovative SMEs'.

Recent literature shows an increasing interest in considering alternative sources of information for predicting Small and Medium Enterprises (SMEs) default. Generally, papers in the field use accounting indicators as regressors of bankruptcy prediction models, although it has been demonstrated they do not allow to completely overcome the information opacity that is one of the main barriers preventing SMEs from accessing to credit. Plus, companies balance sheet data are published at least one year late with respect to the reference period, so that they prevent a real time prediction as requested by financial operators. In this paper we propose websites as an additional source of information for forecasting SMEs default. We borrowed this idea from contributions the have used corporate websites to retrieve online proxies of firms’ economic characteristics, such as corporate culture or firm performance. Our work explores the joint use of online and offline data for enhancing correct prediction of default through kernel discriminant analysis. We also study in detail the firms where the combination is successful in order to highlight possible patterns for future applications.

Importance: Understanding the long-term immune response against SARS-CoV-2 infection in children is crucial to optimize vaccination strategies. Although it is known that SARS-CoV-2 antibodies may persist in adults 12 months after infection, data are limited in the pediatric population. Objective: To examine long-term anti-SARS-CoV-2 spike receptor-binding domain (S-RBD) IgG kinetics in children after SARS-CoV-2 infection. Design, Setting, and Participants: In this single-center, prospective cohort study, patients were enrolled consecutively from April 1, 2020, to August 31, 2021, at the COVID-19 Family Cluster Follow-up Clinic, Department of Women's and Children's Health, University Hospital of Padua. A cohort of 252 COVID-19 family clusters underwent serologic follow-up at 1 to 4, 5 to 10, and more than 10 months after infection with quantification of anti-S-RBD IgG by chemiluminescent immunoassay. Exposures: SARS-CoV-2 infection. Results: Among 902 study participants, 697 had confirmed SARS-CoV-2 infection, including 351 children or older siblings (mean [SD] age, 8.6 [5.1] years) and 346 parents (mean [SD] age, 42.5 [7.1] years). Among 697 cases, 674 (96.7%) were asymptomatic or mild. Children had significantly higher S-RBD IgG titers than older patients across all follow-up time points, with an overall median S-RBD IgG titer in patients younger than 3 years 5-fold higher than adults (304.8 [IQR, 139.0-516.6] kBAU/L vs 55.6 [24.2-136.0] kBAU/L, P <.001). Longitudinal analysis of 56 study participants sampled at least twice during follow-up demonstrated the persistence of antibodies up to 10 months from infection in all age classes, despite a progressive decline over time. Conclusions and Relevance: In this cohort study of Italian children and adults following SARS-CoV-2 infection different kinetics of SARS-CoV-2 antibodies were found across several age classes of individuals with asymptomatic or mild COVID-19, which could help in optimizing COVID-19 vaccination

The containment measures following COVID-19 pandemic drastically reduced airway infections, but they also limited the access of patients to healthcare services. We aimed to assess the antibiotic prescription patterns in the Italian paediatric primary care setting before and after the containment measures implementation. For this retrospective analysis, we used a population database, Pedianet, collecting data of patients aged 0–14 years enrolled with family paediatricians (FP) from March 2019 to March 2021. Antibiotic prescriptions were classified according to WHO AWaRe classification. An interrupted time series evaluating the impact of the containment measures implementation on the monthly antibiotic index, on the access to watch index, and on the amoxicillin to co-amoxiclav index stratified by diagnosis was performed. Overall, 121,304 antibiotic prescriptions were retrieved from 134 FP, for a total of 162,260 children. From March 2020, the antibiotic index dropped by more than 80% for respiratory infections. The Access to Watch trend did not change after the containment measures, reflecting the propensity to prescribe more broad-spectrum antibiotics for respiratory infections even during the pandemic. Similarly, co-amoxiclav was prescribed more often than amoxicillin alone for all the diagnoses, with a significant variation in the trend slope for upper respiratory tract infections prescriptions.

As recently stated by EUROSTAT, Small and Medium Enterprises (SMEs) play a crucial role in the economy of European Union (EU). This is particularly true for Italy that has one of the largest share of SMEs in the Euro zone. In such a context, the assessment of the Italian firms creditworthiness is a priority, especially in the early stage of a SMEs life. The aim of this work is to explore the effectiveness of the Feedforward Neural Network for Italian firms defaults prediction in comparison with standard linear models (Logistic and Probit). Our analysis considers the Italian manufacturing SMEs in a short time span (2016-2017). Results, obtained via two different sampling strategies, are not clear-cut.

The objective of credit scoring is to develop accurate rule of classification that aids to distinguish between good and bad clients. In this context, also Statistical Learning (SL) techniques have been explored, for building models that estimate the clients’ probability of insolvency. Although there are some encouraging results in literature, two main issues makes this classification task very hard: (i) high imbalance ratio between the two groups in the target variable and (ii) the effect of hyperparameter settings on overall performance. In this work, Bayesian Optimization (BO) is used to optimize the hyperparameters of a cost sensitive eXtreme Gradient Boosting (XGBoost) model. Experimental results reveal that the proposed solution is a promising starting point for future development

We propose the use of online indicators, scraped from the firms’ websites, to predict default risk for a sample of Spanish firms via nonlinear discriminant analysis and the logistic regression model.

The growth of cultural firms is important in developing local economies, enhancing employment, and improving urban sustainability, but it is difficult to achieve in fragmented industries that are populated by the smallest firms and offer a particularly unfavorable context for growth. The study takes a contingency perspective and contributes to both the literature on business strategy in fragmented industries and that on the growth of small and medium-sized enterprises (SMEs) by identifying the strategies implemented in a fragmented cultural industry, determining which of them are associated with a firm’s expected growth, and finding the few firm-specific factors that are associated with growth. It also complements the extant literature on art galleries by looking at them from the understudied strategic perspective. Suggestions for practitioners and policy makers are included.

Differently from traditional transaction data, social media data are difficult to investigate due to their volume, variety and velocity. Indeed, the knowledge extraction from social media data raises several issues especially for what concerns statistical exploration and synthesis of complex information. Our work aims to study users' preferences, stated on a social media platform, in order to aid businesses to make their marketing communication decisions. We rely our analysis upon 5685 Italian Facebook users interested in pharmaceutical products and health. The data have been collected at the end of 2014 and are focused on Likes actively expressed by the subjects on specific categories of interests (TV Channels and Magazines). Through a factorial analysis we uncovered significant associations between marketing communication Media and users' profiles. This allows sketching out a marketing strategy in twofold actions: first, identifying the target group to reach and, then, the nearest suitable channel where to develop the marketing communication.

This paper addresses the issue of measuring tolerance, viewed as a multifaceted phenomenon involving several different social domains. We develop a multidimensional index for Likert-scale data, characterized by the following features: (i) it reflects the individual’s intensity of tolerant attitudes towards each social domain; (ii) the index can be broken down by dimension in order to determine the contribution of each dimension to overall tolerance; (iii) the index combines the different dimensions of tolerance using a weighted scheme that reflects the importance of each dimension in determining the overall level of tolerance. To show how this new measure of tolerance works in practice, we carry out a case study using an Italian recent survey asking the opinion of university students about different subjects, such as interreligious dialog, women/religion relationship, religion/death relationship, homosexuality, and multicultural society.

The Small Business Act of the European Commission in 2008 acknowledges the key role of Small and Medium Enterprises (SMEs) in the EU economy. This is particularly relevant for Italy, which has the largest share of SMEs in Europe, as well as for other countries such as Portugal, Spain and Greece. On the other hand, SMEs experience more difficulties in their early stages mainly due to high market competition and credit constraints, as highlighted by Fritsch and Weyh (2006). For these reasons, the study of SMEs default risk is always relevant. There are several papers studying firm default factors in a single country (see Ciampi, 2015, Fantazzini and Figini, 2009, Flix and dos Santos, 2018). The literature concentrates mainly on financial indicators built on businesses’ balance sheets, which are available about two years late with respect to their reference period. This diminishes the significance of the results, both for credit risk and policy aims, and particularly in a forecasting perspective. The purpose of this paper is to provide a preliminary study on a sample of spanish firms selected from the SABI, Sistema de Análisis de Balances Ibéricos, which is listed among Bureau van Dijk databases. The analysis will be carried out according to both parametric and non-parametric discrimination techniques, with the standard construction of a training set on which to build a model and a validation set to test the validity and robustness of the results, and, in the end, the reliability of the model in predicting default. Finally we present a new proposal: a scheme to understand to what extent firms’ default can be predicted by substituting the traditional data sources (offline information) with data collected from their corporate websites (online information) in order to exploit more up-to-date information.

The objective of quantitative credit scoring is to develop accurate models of classification. Most attention has been devoted to deliver new classifiers based on variables commonly used in the economic literature. Several interdisciplinary studies have found that personality traits are related to financial behaviour; therefore, psychological traits could be used to lower credit risk in scoring models. In our paper, we considered financial histories and psychological traits of customers of an Italian bank. We compared the performance of kernel-based classifiers with those of standard ones. We found very promising results in terms of misclassification error reduction when personality attitudes are included in models, with both linear and non-linear discriminants. We also measured the contribution of each variable to risk prediction in order to assess importance of each predictor

In recent studies, there has been a growing interest toward tolerance and its implications in social and economic systems. So far, the openness of people to homosexuals or to foreign-born people has been considered the best indicator of tolerant attitudes. In this paper, we consider tolerance as a multi-faced phenomenon involving several different social domains. The aim is to provide some recommen- dations on how to develop a multidimensional index for tolerance in the case of survey’s items measured by Likert-scale. Our discussion relies on a case study about a student survey carried out at Milan-Bicocca University.

Credit scoring models are generally trained using customers credit history and demographics. Recent works in interdisciplinary studies have showed that alternative source of information, such as psychological traits or behavioural attitudes, can aid to improve default prediction. In our work, we would like to verify if an explorative segmentation based on financial knowledge, preferences and personality traits is able to detect different customer typologies also in terms of financial credit performance. This could be the case when the bank management has to deal with new clients or with those without a long banking history. The segmentation has been derived employing Hierarchical clustering on Factors coming from non-linear Principal Component Analysis (PCA). The Kernel-PCA allowed to map data indirectly in a very-high-dimensional space F where is simple to construct a hyperplane that divides the points into arbitrary clusters. The choice of the kernel functions and its parameters provided different kernel factors which produced, in turn, alternative clusters solutions. All the partitions have been ranked using alternative criteria that measure aspects of clusters validity.

Big Data are hard to capture, store, search, share, analyze, and visualize. Without any doubts, Big Data represent the new frontier of data analysis, although their manipulation is far to be realized by standard computing machines. In this paper, we present a strategy to process and extract knowledge from Facebook data, in order to address marketing actions of a pharmaceutical company. The case study relies on a large Italians sample, interested in wellness and health care. The results of the study are very sturdy and can be easily replicated in different contexts.

In this paper we explore the relationship between the Industrial Production Index (IPI), the confidence index for the manufacturing sector and its sub-indexes and Google searches for several words linked to the economic situation, for the period January 2004 - September 2016 on Italian data. Significant correlations between the selected indicators point to probable comovements of same. Adding one observation at a time since the first forewarning signs of the 2008 crisis, we find that a few Google searches and the IPI cointegrate, particularly during the strong downward trend leading to January 2009, while no confidence indicators cointegrate with the IPI. These findings suggest that concern about economic conditions expressed through searches in google and the IPI or the confidence indexes are influenced by common circumstances. Recursive forecasts of the IPI through VECM models suggest that the evolution of the IPI can be well mimicked using the real time Gtrends selected variables.

L’analisi dell’evoluzione della soddisfazione nel contesto di business, è di solito basato su studi pseudo-panel, perchè sono meno costosi e facili da costruire con i dati disponibili. Come in casi cross-section, informazioni dettagliate sui clienti vengono raccolte in ogni istante temporale, ma il confronto dinamico avviene generalmente considerando soli pochi ritardi temporali (a causa del breve ciclo di vita di prodotti e servizi). Di conseguenza, nel nostro contributo proponiamo l’utilizzo dell’analisi fattoriale multipla duale. Tale tecnica permette la sintesi delle tabelle multivariate e la visualizzazione delle stesse su uno spazio comune che fa luce su traiettorie di soddisfazione dei clienti. I vantaggi della tecnica vengono illustrati in un caso di studio relativo ad una banca italiana, The analysis of the evolution of satisfaction in business context is usually based on pseudo panels studies, because they are less costly and easy to build with the available data. As in the cross-section case, detailed information about customers are collected at each time point, but the dynamic comparison generally involves few temporal lags (due to short life time of products and services). Accordingly, in our paper we apply the Dual Multiple Factor Analysis. Such a technique allows the synthesis of the multivariate tables and their visualization on a common space that sheds light on customers’ trajectories of satisfaction. A real case study of an Italian bank is illustrated

The objective of quantitative credit scoring is to develop accurate models that can distinguish between good and bad applicants (Baesens et al, 2003). Statistical research has focused on delivering new classification methodologies based on neural networks or support vector machines that provide better predictions respect to standard classifiers but poor interpretation. A step forward in such task, it is represented by linear reconstruction of kernel discriminant proposed by Liberati et al (2015) which combines the effective classification results, due to application of no-linear functions, with an easy interpretability of the data. In reality, insolvency, is not always a negative occurrence but the opportunity to generate more profit for a financial institution. Indeed, it can be considered a marketing leverage to reinforce customers’ loyalty. In this paper, we approach credit risk modeling into this perspective. We compare the classification solution that focuses on prediction performance with an explanatory regression as Tobit that models both default probability and duration. Results will be illustrated in a double perspective (risk management and customers' segmentation), in order to identify which covariates affect the insolvent behaviour the most.

The association between service quality and customer satisfaction represents one of the fundamental relationships in marketing research. Although cross-sectional analysis has been commonly applied in such a context, the shift to a more dynamic approach appears to be a worthwhile change. A broad range of statistical methods exist for the management of data that arises from longitudinal designs. Based on pseudo-panel data collected from the banking sector, a latent growth curve model is proposed for the analysis of service quality in order to capture both intra-individual and inter-individual changes in customer satisfaction over time. Results also describe the impact of reliability on overall satisfaction for different customers’ profiles

Due to the recent financial turmoil, a discussion in the banking sector about how to accomplish long term success, and how to follow an exhaustive and powerful strategy in credit scoring is being raised up. Recently, the significant theoretical advances in machine learning algorithms have pushed the application of kernel-based classifiers, producing very effective results. Unfortunately, such tools have an inability to provide an explanation, or comprehensible justification, for the solutions they supply. In this paper, we propose a new strategy to model credit scoring data, which exploits, indirectly, the classification power of the kernel machines into an operative field. A reconstruction process of the kernel classifier is performed via linear regression, if all predictors are numerical, or via a general linear model, if some or all predictors are categorical. The loss of performance, due to such approximation, is balanced by better interpretability for the end user, which is able to order, understand and to rank the influence of each category of the variables set in the prediction. An Italian bank case study has been illustrated and discussed; empirical results reveal a promising performance of the introduced strategy.