Blog

Comparing Traditional Time Series Models and Deep Learning for Load Forecasting: SARIMAX vs. NBEATSx in the Swiss Power Grid

Machine Learning
Python
Energy Economics
Time Series
This is my semester thesis on day-ahead load forecasting, comparing advanced models like NBEATSx and LightGBM with traditional approaches such as SARIMAX. The findings reveal surprising trade-offs between predictive accuracy and economic impact, particularly during holidays and peak periods. Key insights highlight the potential of ensemble methods and the importance of aligning loss functions with economic consequences.
Dec 14, 2024
Mathias Steilen

Classical vs. Heuristic Optimisation: Comparing MILP, Genetic Algorithms and Simulated Annealing for Pumped Storage Hydropower Optimisation

Optimisation
Python
Energy Economics
I implement the genetic algorithm and simulated annealing to optimise the operation of a pumped storage system in the day-ahead market. How do they compare with the solution found by classical mixed-integer linear programming? Let’s find out.
Sep 20, 2024
Mathias Steilen

Forecasting Swiss Electricity Demand

Machine Learning
R
Energy Economics
The goal of this project is to test the ability of supervised machine learning models, namely elastic net, gradient boosting and random forest, to predict end-user electricity consumption in Switzerland.
Mar 9, 2023
Mathias Steilen

Schweizerische Wasserkraft

Data Visualisation
R
Visualisierungen über die Standorte, Arten und Füllungsgrade der Wasserkraft in der Schweiz mit Daten des Bundesamts für Energie
Dec 15, 2022
Mathias Steilen

Luxembourgish Student Aid

Data Visualisation
R
Analysing country level data of Luxembourgish financial aid for students, which gives insights into where students go, what they study and what subjects are prevalent by sex
Nov 14, 2022
Mathias Steilen

Using Classification Models and Gain Curves For Profit Maximisation

Machine Learning
R
Optimisation
This post shows how customer churn classification models can be set up, trained and used to create financial value within organisations, specifically using gain curves and classification thresholds.
Oct 25, 2022
Mathias Steilen

Stacking Models To Predict Airbnb Prices In Manhattan (NYC)

Machine Learning
R
The purpose of this post is to demonstrate the usefulness of the package stacks in blending individual machine learning models together into a linear combination of them (using LASSO regression), often increasing final model performance.
Oct 7, 2022
Mathias Steilen

Deploying A Model Predicting Used Car Prices With Shiny

Machine Learning
Web Scraping
R
The data for this post was scraped from a Swiss online platform. The goal of this post is to demonstrate a deployed model with Shiny, which allows users to enter information about their car and get back a prediction for the price in Swiss francs. A bonus is using tokenisation for extracting predictive information from the natural language title of each listing.
Sep 23, 2022
Mathias Steilen

Scraping and Modelling Used Motorcycles

Machine Learning
Web Scraping
Data Visualisation
R
This blog post is about how to create your own data sets by scraping websites with rvest. Specifically, I will scrape all available used motorcycles from a Swiss online platform, explore the data, make conclusions and train a model on the prices.
Sep 19, 2022
Mathias Steilen

Birthday Problem

Simulation
R
The birthday problem is the surprising fact that in a group of just 23 people, there’s a 50-50 chance that two of them share the same birthday, which is way more likely than expected by most.
Aug 9, 2022
Mathias Steilen

US Wind Energy

Data Visualisation
Energy Economics
R
Visualising the wind turbine data base from USGS, which covers 71,666 wind turbines in the United States over a period from 1981 to 2022.
Aug 9, 2022
Mathias Steilen

Swiss End-User Electricity Consumption

Data Visualisation
Energy Economics
R
Visualising Swiss End-User Electricity Consumption data provided by Swissgrid over the past decade
Aug 9, 2022
Mathias Steilen

Alternative Fuel Stations in the US

Data Visualisation
R
Visualising alternative fuelling stations in the US
Jul 29, 2022
Mathias Steilen

The St. Petersburg Paradox

Probability
R
The St. Petersburg paradox is a puzzling situation where a simple gambling game seems to promise a lot of money, but most people wouldn’t be willing to pay much to play it, even though the math suggests they should.
Jul 16, 2022
Mathias Steilen

Do Known Mutual Fund Brands Outperform?

Data Visualisation
Finance
R
Analysing if mutual funds from better known fund managers performed better as opposed to mutual funds from lesser known fund managers
Jul 8, 2022
Mathias Steilen

Do Moving Averages Convey Information About Stock Price Movements?

Machine Learning
Finance
R
There is plethora of day trading coaches claiming technical indicators like moving averages are useful for timing the market in stock trading. With this short post, I want to investigate whether supervised machine learning models like random forests are capable of extracting information from moving averages (technical indicator) about the future direction of a stock on a daily basis.
Jun 30, 2022
Mathias Steilen

Application Of Machine Learning Methods To Mutual Fund Selection

Machine Learning
Finance
R
This is an executive summary of my bachelor’s thesis on machine learning in mutual fund selection. The main findings and implications with regard to practice are presented.
Jun 3, 2022
Mathias Steilen

How Often Should You Check Your Portfolio?

Probability
Finance
R
Once you have invested in the stock market, it is hard not to check your portfolio every morning. But how often should you check it to maximise your mental health? The quick answer is: As infrequently as possible.
Jun 2, 2022
Mathias Steilen

The Monty Hall Dilemma

Probability
R
The Monty Hall problem is a brain-teasing scenario in which you’re asked to choose between three doors, with a hidden prize behind one of them, and switching your choice after one door is revealed can dramatically increase your chances of winning, which might sound counterintuitive at first.
May 31, 2022
Mathias Steilen
No matching items