Big data - Data mining - Analyse prédictiveAnalyse de Données pour la Qualité (ERE)

Knowledge of the application of multidimensional data analysis methods to data mining To do this, it is necessary to: Know: • Data usage issues for businesses • The environmental impacts of artificial intelligence and digital technology • Different types of databases and open data resources • Integrating Big Data into business strategy Master: • The CRISP data analysis methodology and its implementation • Validation of the quality of data used • A tool (Knime) to explore, analyze data • Basic quantitative models and methods for segmenting observations or predicting a phenomenon (K-means algorithm, linear or logistic regression, decision trees) • Validation and interpretation of results.

The chapters being tested in TD are marked ( • General presentation of data science • Environmental impacts of digital and artificial intelligence • Historical context: from academic research to business • Life cycle of a data recovery project ( CRISP methodology) • Project framing: essentials • Collects internal business data, and external data (open-data) • Exploring data: understand their quality and relevance, select and filter data, introduce tools like Knime. • Modeling: segmentation, detection of anomalies • Prediction: classification, regression, reduction of bias and variance of a model, automatic selection of variables • Statistical validation of models () : a dilemma of bias-variance/complexity, over-learning, cross-validation, study of model predictions • Interpretation of models: simple / complex models • Restitution and measurement of return on investment in enterprises: visualisation of results, explanation of the work done, measurement of the impact for the enterprise

Table control (Duration 1h)