Data Mining Practical Machine Learning Tools and Techniques

Peer-to-peer (P2P) lending demands effective and explainable credit risk models. Typical machine learning algorithms offer high prediction performance, but most of them lack explanatory power. However, this deficiency can be solved with the help of the explainability tools proposed in the last few years, such as the SHAP values. In this work, we assess the well-known logistic regression model and several machine learning algorithms for granting scoring in P2P lending. The comparison reveals that the machine learning alternative is superior in terms of not only classification performance but also explainability. More precisely, the SHAP values reveal that machine learning algorithms can reflect dispersion, nonlinearity and structural breaks in the relationships between each feature and the target variable. Our results demonstrate that is possible to have machine learning credit scoring models be both accurate and transparent. Such models provide the trust that the industry, regulators and end-users demand in P2P lending and may lead to a wider adoption of machine learning in this and other risk assessment applications where explainability is required.

Explainability of a Machine Learning Granting Scoring Model in Peer-to-Peer Lending

Peer to Peer (P2P) Lending Problems and Potential Solutions: A Systematic Literature Review

This paper presents a benchmark of supervised Automated Machine Learning (AutoML) tools. Firstly, we analyze the characteristics of eight recent open-source AutoML tools (Auto-Keras, Auto-PyTorch, Auto-Sklearn, AutoGluon, H2O AutoML, rminer, TPOT and TransmogrifAI) and describe twelve popular OpenML datasets that were used in the benchmark (divided into regression, binary and multi-class classification tasks). Then, we perform a comparison study with hundreds of computational experiments based on three scenarios: General Machine Learning (GML), Deep Learning (DL) and XGBoost (XGB). To select the best tool, we used a lexicographic approach, considering first the average prediction score for each task and then the computational effort. The best predictive results were achieved for GML, which were further compared with the best OpenML public results. Overall, the best GML AutoML tools obtained competitive results, outperforming the best OpenML models in five datasets. These results confirm the potential of the general-purpose AutoML tools to fully automate the Machine Learning (ML) algorithm selection and tuning.

/pdf/a-comparison-of-automl-tools-for-machine-learning-deep-33myfrok54.pdf

A Comparison of AutoML Tools for Machine Learning, Deep Learning and XGBoost

Learning nonstationary data streams has been well studied in recent years. However, most of the researches assume that the class imbalance of data streams is relatively balanced. Only a few approaches tackle the joint issue of concept drift and class imbalance due to its complexity. Meanwhile, the existing chunk ensembles for classifying imbalanced nonstationary data streams always need to store previous data, which consumes plenty of memory usage. To overcome these issues, we propose a chunk-based incremental ensemble algorithm called Dynamic Updated Ensemble (DUE) for learning imbalanced data streams with concept drift. Compared to the existing techniques, its merits are five-fold: (1) it learns one chunk at a time without requiring access to previous data; (2) it emphasizes misclassified examples in the model update procedure; (3) it can timely react to multiple kinds of concept drifts; (4) it can adapt to the new condition when switching majority class to minority class; (5) it keeps a limited number of classifiers to ensure high efficiency. Experiments on synthetic and real datasets demonstrate the effectiveness of DUE in learning nonstationary imbalanced data streams.

Incremental learning imbalanced data streams with concept drift: The dynamic updated ensemble algorithm

Due to the growth of data and widespread usage of Machine Learning (ML) by non-experts, automation and scalability are becoming key issues for ML. This paper presents an automated and scalable framework for ML that requires minimum human input. We designed the framework for the domain of telecommunications risk management. This domain often requires non-ML-experts to continuously update supervised learning models that are trained on huge amounts of data. Thus, the framework uses Automated Machine Learning (AutoML), to select and tune the ML models, and distributed ML, to deal with Big Data. The modules included in the framework are task detection (to detect classification or regression), data preprocessing, feature selection, model training, and deployment. In this paper, we focus the experiments on the model training module. We first analyze the capabilities of eight AutoML tools: Auto-Gluon, Auto-Keras, Auto-Sklearn, Auto-Weka, H2O AutoML, Rminer, TPOT, and TransmogrifAI. Then, to select the tool for model training, we performed a benchmark with the only two tools that address a distributed ML (H2O AutoML and TransmogrifAI). The experiments used three real-world datasets from the telecommunications domain (churn, event forecasting, and fraud detection), as provided by an analytics company. The experiments allowed us to measure the computational effort and predictive capability of the AutoML tools. Both tools obtained high-quality results and did not present substantial predictive differences. Nevertheless, H2O AutoML was selected by the analytics company for the model training module, since it was considered a more mature technology that presented a more interesting set of features (e.g., integration with more platforms). After choosing H2O AutoML for the ML training, we selected the technologies for the remaining components of the architecture (e.g., data preprocessing and web interface).

/pdf/a-scalable-and-automated-machine-learning-framework-to-z55uga5oxp.pdf

A scalable and automated machine learning framework to support risk management

Following the story line of Information and Communication Technologies (ICT), since the first room sized computer to the nowadays finger sized processors, there is an uncontrollable increase of the number of devices per person, each time more powerful, as well as a massive access to the Internet. Altogether, this technological movement brought a new paradigm referred to as IoT (Internet of Things), that is quickly affecting almost all dimensions of life, supported on complex (ICT) infrastructures whose safety and security properties need to be addressed carefully. Although in a first look the Cloud architecture showed itself able to handle the requirements, currently this assumption is no longer true. The exponential growth of devices with limited resources exploited some limitations of the Cloud, namely the lack of bandwidth. This happened because, by design, it was architected to deal with high-level data produced by computer-like terminals. Since the introduction of the IoT paradigm, where billions of devises are continuously uploading minimal logs, there was a need to introduce a new middle-ware (at the edge), capable of gathering all elementary information, producing high-level information adequate to the Cloud. This project aimed to design such an edge, with specific requirements to support Smart Cities or similar environments. Following the generic Edge Computing paradigm, the architecture Edge4All currently aims to: accept and deal with different devices security levels; dynamically process the information and provide it through an API; keep historical data to provide a proper big data/machine learning information source; offer a platform to securely manage the deployed devices; and design a dynamic and intuitive dashboard displaying the information, warnings and alert messages, aiming to keep both the regular user well informed and offer an efficient decision-making tool, to related decision makers.

/pdf/edge4all-edge-platform-for-smart-city-3fmui4zmai.pdf

EDGE4ALL - Edge Platform For Smart City

The financial market is one of the major consumers of data mining techniques, and the main reason is their efficiency to analyze complex data. One important trait shared between most financial applications is class imbalance. Since traditional classification methods assume nearly balanced classes and equal misclassification costs, they usually fail to deal with imbalanced data. However, in financial contexts, problems are usually imbalanced, and instances from the minority class are known for deficits of millions of dollars every year, e.g., credit card frauds, money laundering transactions and so forth. Over the years, several techniques for dealing with class imbalance have been developed, such as sampling techniques and algorithm adaptations. In this study, we analyze how different sampling techniques impact the performance of different classification systems on financial applications. Results show that, for the given datasets, sampling techniques allow the improvement of prediction performance of the minority class while also improving overall classification rates. Nevertheless, their use often deteriorates the performance in predicting the majority class.

An Experimental Perspective on Sampling Methods for Imbalanced Learning From Financial Databases

Objective: To describe how organizational culture influenced on elaboration process of participatory strategic planning (PEP) of a multicampi state university in Brazil’s Midwest region.Methodology/approach: A qualitative approach was adopted with the Ex-post facto case study method, with semi-structured interviews and participant direct observation. The analysis was based on the pattern-matching technique and narrative analysis, from the perspective of two constitutive elements of analysis: (i) Strategy as a collective process and (ii) PEP as an instrument for strengthening organizational culture.Originality/Relevance: Educational institutions are considered complex, which hinders the applicability of tools and management commonly used in other types of organizations. The gap of this study was the combination of researching organizational culture together with the theme of strategic planning in a higher education institution and thereby expanding research on university management.Main results: It was verified that the interaction provided by the process of discussion and construction of PEP contributed to the strengthening of the organizational culture, enabled individuals to internalize organizational values and beliefs through a process of socialization, acculturation, and social integration among the various actors involved.Theoretical/methodological contributions: It highlight the importance of strategic planning in different organizational cultures, especially in higher education institutions. The methods adopted contributed to confirm the need to use various sources for the necessary deepening of the development of a case study.Social contributions / for management: Strengthening the organizational culture.

/pdf/analise-da-influencia-da-cultura-organizacional-no-4tehs8xmit.pdf

Análise da influência da cultura organizacional no planejamento estratégico participativo de uma universidade estadual multicampi na região centro-oeste do brasil

The project TheRoute (Tourism and Heritage Routes including Ambient Intelligence with Visitants' Profile Adaptation and Context Awareness) aim is to conduct studies, research and experimentation around the challenge of automatic generation of routes for visitors. The suggested routes fit the profile of visitors and groups of visitors, including aspects like emotion/mood/personality, and aware of the context (e.g. weather, security). The focus region is north of Portugal, developed around a location (urban or rural) or path (e.g. Portuguese Santiago Route or Douro) or around a theme (e.g. musician, writer). Points of Interest (POI) will be modeled. Also, the system is able to use different platforms for the tourist (desktop, notebook, tablet, Smartphone). TheRoute has been developed considering the notion of Ambient Intelligence (AmI) and Context Awareness. It consists of a context-aware touristic recommendation system which can suggest a series of locations to a user based on their personality, preferences, and location. This paper focuses on the proposed route generation processes, including examples of the overall resulting behavior. We discuss two proposed algorithms (an iterative branch-and-bound based process & sequential addition of the nearest feasible sub-route) for route generation, considering contextual model designing based on user's psychology, and point of interest and availability. The route generation tests that were performed included some of the most influential Porto points of interest.

Luís Ferreira

Papers

A scalable and automated machine learning framework to support risk management

EDGE4ALL - Edge Platform For Smart City

An Experimental Perspective on Sampling Methods for Imbalanced Learning From Financial Databases

Análise da influência da cultura organizacional no planejamento estratégico participativo de uma universidade estadual multicampi na região centro-oeste do brasil

Tourism and Heritage Routes Including Visitants Profile Adaptation and Context Awareness