Machine Learning in Python with Financial and Insurance Applications

Machine Learning in Python with Financial and Insurance Applications
Machine Learning in Python with Financial and Insurance Applications

This course is an indispensable asset if you already have or if you aspire to have a technical career in the financial or insurance industry since nearly every aspect of them is heavily affected by the adoption of Machine Learning (ML). Also, people with managerial or non-technical positions will enormously benefit from this course because grasping the concepts of ML will enable them to make more informative decisions, communicate more effectively with the technical staff and understand in depth the result of their analysis.

The language chosen to apply ML is Python because it is one of the most popular languages in these fields with vast and continuously evolving libraries. No prior knowledge of Python is required. The language is presented in a concise, understandable and quick-to-use way with strong emphasis on ML.

Professor of of Mathematics, NKUA, Greece

The program is addressed to anyone interested in Machine Learning, Data Analysis and Deep Learning techniques and their practical applications in business, with an emphasis in financial and insurance applications.

The course aims at introducing the participants to modern Machine and Deep Learning techniques (MDL), with the objective of them acquiring unique skills on the field and becoming competitive in the international job market. The participants will get to know in a systematic way MDL, with real applications from many business sectors. Intuitive learning will be combined with rigorous presentation of the models and techniques, with the goal of constructing practical codes and solving real-life problems needed in the industry. Although primary focus will be placed on financial and insurance applications (bearing in mind that the experience gained is easily transferable to various sectors and industries),many more day to day, real-life applications of MDL, like image processing, will also be covered, so that participants gain a more complete view of the general area. The programming language used will be Python. Participants who successfully complete this course will be able to enhance their careers as data analysts, quantitative analysts, risk analysts, modelers or decision makers in the financial, insurance, banking or in any sector where data analysis and predictions play an important role in decision making. Anyone who wants to acquire a compact, practical and easily applicable knowledge on this field will benefit tremendously from this course.

This course starts from the very basics aspects of ML and continues to the most modern and advanced topics. Expose to the material is made step-by-step in a systematic, structured and understandable way. The concepts are explained from an intuitive point of view and strong emphasis is given to applications. By the end of this course, you will be able to analyze from scratch the most popular problems in finance and insurance, to apply ML models in Python to tackle them and to interpret the results. The “hands-on” and “do-it-yourself” philosophy is adopted throughout the course so that you can benefit the most by applying yourself ML to practical problems in Finance and Insurance. 
The most modern techniques in the area of EDA, supervised unsupervised, unsupervised learning and neural networks are demonstrated. All the relevant libraries in Python are explained and secrets of how to use them efficiently are demonstrated.
Each one of the 11 modules is self-contained, easy to understand and is accompanied with a number of labs where you have the opportunity to apply yourself the various models. Labs also contain summary the theory, exercises and comments on the solutions and practical tips for use.The last module contains real life financial and insurance applications so that you can have first-hand experience of how to approach problems in these two areas. 
Course Outline:
1.    “Machine Learning Introduction”: Artificial Intelligence, Machine Learning, Supervised and Unsupervised Learning, Deep Learning, Machine Learning projects, Types of Features and Variables, Categorical Variables, Numeric Variables, Exploratory Data Analysis, Feature Engineering, Collecting Insights from models, Machine Learning Models Structure and Formalization. Number of Coding Labs: 3.
2.    “Data Cleaning and Manipulation”: Exploratory Data Analysis, One Hot Encoding, Label Encoding, Feature Transformation, Feature Scaling, Normalization, Standardization, Handling with Missing Data, Handling Outliers. Number of Coding Labs: 3.
3.    “Regression part 1”: Supervised Learning, Simple Linear Regression, Multiple Linear Regression, Polynomial Regression, Train-Test Split approach for evaluation of Machine Learning models and evaluation metrics for regression models. Number of Coding Labs: 3.
4.    “Regression part2”: Regularization, Overfitting, Underfitting, Error in Machine Learning Models, Bias-Variance Tradeoff, Ridge Regression, Lasso Regression, Elastic Net Regression, Hyperparameters in Machine Learning Models, Cross Validation approach in model evaluation. Number of Coding Labs: 2.
5.    “Classification part1”: Supervised Machine Learning, Classification Problems, Logistic Regression, Support Vector Classifier,Hyperparameters, evaluation metrics for classification models. Number of Coding Labs: 2.
6.    “Classification part2”: Supervised Machine Learning, Classification Problems, k-Nearest Neighbors Classifier (k-NN), Distance Metrics in Machine Learning, Distance-Based Models, Decision Tree Classifier, Random Forest Classifier, Naïve Bayes Classifier. Number of Coding Labs: 3.
7.    “Unsupervised Learning”: Unsupervised Machine Learning, Clustering, Dimensionality Reduction, Curse of Dimensionality, Pattern Recognition, k-Means Clustering, Elbow method, DBSCAN clustering, Silhouette Score for Clustering evaluation, Principal Components Analysis, Factor Analysis. Number of Coding Labs: 3.
8.    “Neural Networks-Deep Learning part1”: Deep Learning, Neural Networks, Artificial Intelligence, Types of Neural Networks, Artificial Neural Networks, Regression, Classification, Neurons, Layers, Activation Functions, Training of ANNs, Forward Propagation, Backward Propagation, Loss Function, weight update, optimizers, Gradient Descent, Stochastic Gradient Descent, Mini-Batch Stochastic Gradient Descent, Convex and non-convex functions, Adam Optimizer. Number Of Coding Labs: 3.
9.    “Neural Networks-Deep Learning part2”: Activation andCost Functions, Mean Squared Error, Mean Absolute Error, Hubber Loss, Cross Entropy loss, Multi-class Cross Entropy, Chain Rule of differentiation, Problems in Neural Network training, Vanishing Gradient Problem, Exploding Gradient Problem, Dead Neuron Problem, Weight Initialization Strategies, Dropout Techniques. Number of Coding Labs: 3.
10.    “Neural Networks-Deep Learning part3”: Deep Learning, Neural Networks, Recurrent Neural Networks (RNNs), Training of RNNs, Applications of RNNs, Sequential Data, Problems in RNN training, Forward Propagation in RNNs, Backward Propagation in RNNs, LSTM-RNN Networks. Number of Coding Labs: 2.
11.    “Ensemble Machine Learning”: Supervised Machine Learning, Regression, Classification, Overfitting Ensemble Learning, Parallel Models, Sequential Models, Bagging Models, Boosting Models, Decision Tree Regressor, Random Forest Regressor, Extreme Gradient Boosting, XGBoost Regressor, XGBoost Classifier. Number of Coding Labs: 2.
12.    “Applications of ML in Finance and Insurance with Python”: Real-world Machine Learning and Deep Learning cases from Finance, Actuarial Science and in general Insurance. Number of Coding Labs: 4.

A variety of assessment methods are employed, including exams with multiple choice questions, written essays, and Q/A during the live-streaming lectures. These may differ for each individual unit.

Online and distance training learning at National and Kapodistrian University of Athens offers a new way of combining innovative learning and training techniques with interaction with your tutor and fellow trainees from around the world.

The e-learning course is implemented via a user-friendly educational platform adjusted to the Distance Learning Principles. Courses are structured as weekly online meetings; interaction with the course tutor and other trainees takes place in a digital learning environment. The courses are designed to fit around your schedule; you access the course whenever it is convenient for you, however within the given deadlines.

The whole world becomes your classroom as e-learning can be done on laptops, tablets and phones as a very mobile method. Learning can be done on the train, on a plane or even during your trip to Greece!

The educational platform is a portal that offers access to electronic educational material based on modern distance learning technologies. The computer based nature of training means new technology is being introduced all the time to help trainees engage and learn in a tailored way that will meet their needs. E-learners have access to the educational platform with their personal code number in order to browse all relevant training material and interact with their instructors.

Moreover, an online communication system through own personal e-mail account is available in order to make the process easier and more interactive. Trainees can contact directly their tutors or the administration office of the course and share any concerns or anxieties related to the course in order to make the most of their experience.

Every week e-learners are provided with the relevant material, delivered either in the form of video-lectures, text notes and relevant presentations or as a combination of them. The educational material of the course is uploaded gradually, per educational unit. During the course, important info for the smooth conduct of the educational process, such as timetables for the submission of the exercises are announced on the Announcement section of the platform.

For successful completion of the course the e-learner should have fulfill her/his academic obligations, meaning should have submitted all corresponding assessment exercises and have achieved at least an average of 50% grade in the corresponding tests for each module. The score scale ranges from 0 to 100%. Finally, if the total score on one or more lessons of the course does not exceed 50%, trainees can ask for reassessment.

During the course trainees will be attending a training experience designed by academics and lecturers from the National University of Athens as well as from other Universities, Research Institutes and Cultural organizations around Greece.

Interactivity, flexibility and our long tradition guarantee that learning with us offers a successful and rewarding experience. Finally, access to a large variety of material and online resources available in each unit aims to excite your curiosity and guide you in exploring further your favourite topic. Part of the online material can be downloaded providing the chance to quickly refresh your memory after the completion of the course.