Thierry Moudiki

This study examines climate-related financial risk issues, and introduces methods for estimating carbon beta and carbon Shap-ley values. The carbon beta measures how stock returns respond to a portfolio called Brown Minus Green (BMG), that holds a long position in less climate-friendly (brown) stocks and a short position in climate-friendly (green)...

The martingale hypothesis asserts that the best approximation of the future value of a time series, based on its historical values, is its present value. In this document, I propose a statistical approach to test this hypothesis by employing a multiple linear regression model on lagged values, computing the Fisher F-statistic, and assessing station...

This study examines climate-related financial risk issues, and introduces methods for estimating carbon beta and carbon Shap-ley values. The carbon beta measures how stock returns respond to a portfolio called Brown Minus Green (BMG), that holds a long position in less climate-friendly (brown) stocks and a short position in climate-friendly (green)...

This paper presents a gradient boosting machine inspired by the LS Boost model introduced in [Friedman, 2001]. Instead of using linear least squares models as base learners, the proposed algorithm is model-agnostic and employs randomized neural networks. The combination of gradient boosting and randomized neural network base learners provides a pow...

Probabilistic Forecasting with nnetsauce (using Density Estimation, Bayesian inference, Conformal prediction and Vine copulas)

Probabilistic Forecasting with Quasi-randomized networks (and Python package nnetsauce)

This paper introduces deep Quasi-Randomized neural Networks (QRN) for supervised Machine Learning (ML) classification. Whereas classical neural networks rely on backpropagation (gradient descent of loss functions) for their calibration, QRNs are characterized both by quasi-randomized layers of engineered features, and by a classifier, a base model...

Predictive simulation of time series data is useful for many applications such as risk management and stress-testing in finance or insurance, climate modeling, and electricity load forecasting. This paper proposes a new approach to uncertainty quantification for univariate time series forecasting. This approach adapts split conformal prediction to...

Univariate and Multivariate time series (MTS hereafter) are collections of sequential data points observed at different timesteps for measurable indicators. Real-world examples of MTS include-among many others-the monthly totals of international airline passengers from 1949 to 1960, or the daily closing prices of major European stock indices from 1...

Regression-based classifiers using quasi-randomized nnetworks

Python packages for data analytics and Machine Learning

Gradient boosted randomized *neural* networks

Given the increase in relative complexity of Statistical/Machine Learning models – whose billions of parameters are also regularized in someway to prevent overfitting –, and since in certain industries (e.g healthcare,banking, etc.) customers would like to understand how model predictions about them are formed, interpretability is becoming an impor...

This paper is about AdaOpt, a probabilistic multivariable optimization model for statistical/machine learning classification.

This paper is about nonlinear ridge regression models (for regression, both in a frequentist and a Bayesian way, and classification) having a quasi-randomized hidden layer, and two shrinkage parameters.

This preprint is about a tool for stochastic simulation in Finance, Insurance, Economics or Physics (...).

This paper contributes to the development of quasi-randomized networks; neural networks with quasi-randomized hidden layers. It deals in particular with multinomial logistic regression, a supervised learning method that allows to classify statistical/machine learning model observations in multiple categories. The model presented here notably takes...

Multivariate time series forecasting in nnetsauce

This paper contributes to a development of Randomized Neural Networks, and more specifically to a development of the Stochastic Configuration Networks (SCNs). We present a family of learning algorithms based on the SCNs and on ensembles of single layer feedforward networks (SLFNs). They are close to Gradient Boosting, and to Matching Pursuit algori...

This paper contributes to adding a Bayesian Quasi-Random Vector Functional Link network (BQRVFL) to the Machine Learning practitioner’s toolbox. The BQRVFL is an hybrid penalized regression/neural network model, that takes into account input data’s heterogeneity through clustering. Its regression coefficients are governed by a Multivariate Gaussian...

The Own Risk Solvency and Assessment (ORSA) is a set of processes defined by the European prudential directive Solvency II, that serve for decision-making and strategic analysis. In the context of ORSA, insurance companies are required to assess their solvency needs in a continuous and prospective way. For this purpose, they notably need to forecas...

We are interested in obtaining forecasts for multiple time series, by taking into account the potential nonlinear relationships between their observations. For this purpose, we use a specific type of regression model on an augmented dataset of lagged time series. Our model is inspired by dynamic regression models (Pankratz 2012), with the response...

Multivariate time series forecasting using quasi-randomized nnetworks

Yield curve interpolation and extrapolation using no arbitrage short rates models

A discount curve for Insurance Risk Management with exact fit and parsimonious forecasts

The projection of economic and financial risk factors is a key element of prospective analyzes made by life insurers, both for the calculation of reserves under Solvency 2 and for the asset allocation and management of financial risks. This projection is achieved in practice through “economic scenario generators” (ESG), which are inputs for the cal...

Swap curve for insurance pricing

Yield curve interpolation and extrapolation

Diffusion models for insurance valuation

Economic Scenarios Generators for Insurance

Yield curve interpolation and extrapolation; impact on Best Estimate Liabilities reserves