LearnableLoopAI.com – Portfolio of Projects


(Click on a tag below a title, or, click on a category on the far right.)

We assist with your Data, Analytics, Machine Learning, and AI needs, including the development of models of your business operations. But then we go further – we close a learnable loop around your models by measuring your system’s output (instrumentation), sending the output through our agents, and finally prescribing some actions (actuation) to align with your objectives. The end result –- an ongoing more optimized plan of action for your business operations.

The most effective business operations have closed loops. Humans often participate in loops. Sometimes loops are completedly automated. Our delight is to optimize your operations and assist you in making your loops able to learn …

Back to Portfolio of Projects |  LearnableLoopAI.com |  Blog |  LinkedIn


Admin-Website-Consumer Multi-Agent System Using PyMDP (Part 2)

A three-agent system to manage a search ad campaign
Advertizing Industry
Bayesian Inference
Active Inference
Multi-Agent
PyMDP
Python
In Part 1 we setup a three-entity multi-agent system that use PyMDP. In Part 2 we implement the first version of the client’s required proof-of-concept (POC) wherein the…
Nov 29, 2024
Kobus Esterhuysen

Admin-Website-Consumer Multi-Agent System Using PyMDP (Part 1)

Adapting a PyMDP demo to a three-agent system to establish the basic operation of Active Inference
IT Industry
Bayesian Inference
Active Inference
Multi-Agent
PyMDP
Python
In Part 1 we setup a three-entity multi-agent system that use PyMDP. However, we make use of the sequence structure of RxInfer. To do this we modify an existing demo:
Nov 21, 2024
Kobus Esterhuysen

Multi-Agent Collective Intelligence with Active Inference (Part 1)

Considering a two-level structure to derive collective intelligence
Enterprise Industry
Bayesian Inference
Active Inference
RxInfer
Python
In this project the client offers an Enterprise Resource Planning (ERP) system that consists of a number of modules. The system runs as a SaaS (Software-as-a-Service). The…
Oct 14, 2024
Kobus Esterhuysen

Dynamic Trading using Active Inference (Part 2)

Reusing Python code to setup an RxInfer trading system
Investment Industry
Inventory Management
Bayesian Inference
Active Inference
RxInfer
Julia
PythonCall
Python
In Part 2 we switch to Julia and setup the PythonCall and CondaPkg Julia packages to interoperate with the reused Python code. We also collect the code to be reused in a…
Aug 30, 2024
Kobus Esterhuysen

Dynamic Trading using Active Inference (Part 1)

Reusing Python code to setup an RxInfer trading system
Investment Industry
Inventory Management
Powell Unified Framework
Reinforcement Learning
Python
RxInfer
Dynamic optimal investment decisions can increase ROI substantially when applied in the right context. This project lays the groundwork for setting up a Dynamic Trading…
Aug 19, 2024
Kobus Esterhuysen

Julia to Python

Making use of existing Python code from Julia and RxInfer
RxInfer
Julia
PythonCall
Python
While practising Active Inference making use of RxInfer it would be beneficial to sometimes make use of existing code in Python. For example, using existing data generation…
Aug 18, 2024
Kobus Esterhuysen

Visualizing Forney Factor Graphs when using RxInfer (Part 5)

Creating a separate FFGViz module
Bayesian Inference
Active Inference
TikZ
RxInfer
Julia
In this part (Part 5) we:
Aug 16, 2024
Kobus Esterhuysen

Visualizing Forney Factor Graphs when using RxInfer (Part 4)

Automating most of the layout configuration
Bayesian Inference
Active Inference
TikZ
RxInfer
Julia
In Part 3 we added decorations for Constrained Forney Factor Graphs (CFFG) as well as for delta factors. In this part we:
Aug 15, 2024
Kobus Esterhuysen

Visualizing Forney Factor Graphs when using RxInfer (Part 3)

Getting closer to the final format
Bayesian Inference
Active Inference
TikZ
RxInfer
Julia
In Part 2 we provided a configurable layout scheme suitable for Forney Factor Graphs (FFGs). We also moved to using TikZ rather than GraphViz (please see Part 2 for an…
Jul 22, 2024
Kobus Esterhuysen

Visualizing Forney Factor Graphs when using RxInfer (Part 2)

Exploring possibilities using TikZ
Bayesian Inference
Active Inference
TikZ
RxInfer
Julia
In part 1 we laid a foundation for the visualization of Forney Factor Graphs (FFGs) when using RxInfer. In the current part (Part 2) we provide a satisfactory way to layout…
Jul 6, 2024
Kobus Esterhuysen

Visualizing Forney Factor Graphs when using RxInfer (Part 1)

Exploring possibilities using GraphViz
Bayesian Inference
Active Inference
GraphViz
RxInfer
Julia
In this project we lay a foundation for the visualization of Forney Factor Graphs (FFGs) when using RxInfer. Followup projects will refine the approach.
Jul 3, 2024
Kobus Esterhuysen

Sales Forecasting with Time-Varying Autoregressive Models (Part 4)

Forecast real sales data
Retail Industry
Bayesian Inference
Autoregressive (AR) Models
Active Inference
RxInfer
Julia
In Part 3 we did the first forecast beyond the available synthetic data. In this part, we make use of real sales data. As before, we use RxInfer to learn the best order (2)…
Jun 19, 2024
Kobus Esterhuysen

Sales Forecasting with Time-Varying Autoregressive Models (Part 3)

Doing the first forecast
Retail Industry
Bayesian Inference
Autoregressive (AR) Models
Active Inference
RxInfer
Julia
In Part 2 of the Sales Forecasting project we setup the foundations of the design and provided a simple implementation. In Part 3 we do the first forecast beyond the…
Jun 18, 2024
Kobus Esterhuysen

Sales Forecasting with Time-Varying Autoregressive Models (Part 2)

Learning the parameters of an autoregressive process
Retail Industry
Bayesian Inference
Autoregressive (AR) Models
Active Inference
RxInfer
Julia
Forecasting sales volume and sales revenue is an essential part of the operations of a business. In Part 1, we analyzed a paper to understand some principles supporting the…
Jun 14, 2024
Kobus Esterhuysen

Sales Forecasting with Time-Varying Autoregressive Models (Part 1)

Analysis of Podusenko, Kouw, de Vries (2021). Message Passing-Based Inference for Time-Varying Autoregressive Models
Retail Industry
Bayesian Inference
Autoregressive (AR) Models
Active Inference
RxInfer
Julia
The purpose of this project is to forecast sales volume and sales revenue. In order to lay the ground work for an approach based on a bayesian reactive message passing…
Jun 1, 2024
Kobus Esterhuysen

Hidden Markov Model with Control (Part 2)

Control a Roomba using Active Inference and RxInfer
Bayesian Inference
Active Inference
RxInfer
Julia
This is the second part of an effort to add control to the Hidden Markov Model RxInfer example. Eventually this work will be used in a more elaborate scheduling project…
May 20, 2024
Kobus Esterhuysen

Hidden Markov Model with Control (Part 1)

Control a Roomba using Active Inference and RxInfer
Bayesian Inference
Active Inference
RxInfer
Julia
This is the first part of an effort to add control to the Hidden Markov Model RxInfer example. Eventually this work will be used in a more elaborate scheduling project…
May 10, 2024
Kobus Esterhuysen

Watching Paint Dry with Active Inference

Navigate a cart to the optimal location using Bayesian Inference and RxInfer
Specialty Trade Contractors Industry
Bayesian Inference
Active Inference
RxInfer
Julia
The client that approached us for a solution is in the spray-painting business. The objects to spray-paint are of various kinds. In addition, various kinds of paint are used…
May 3, 2024
Kobus Esterhuysen

T-maze Active Inference - Policy Inference (Part 1b)

Navigate a simple T-maze using Bayesian Inference and RxInfer
Bayesian Inference
Active Inference
RxInfer
Julia
The overall purpose of the project is to lay the ground work for another project that makes use of a categorical control space. The purpose of the Part 1b project is to…
Apr 30, 2024
Kobus Esterhuysen

T-maze Active Inference - Policy Inference (Part 1a)

Analysis of Koudahl, van de Laar, de Vries (2023). Realising Synthetic Active Inference Agents, Part I–Epistemic Objectives and Graphical Specification Language
Bayesian Inference
Active Inference
RxInfer
Julia
The overall purpose of the project is to lay the ground work for another project that makes use of a categorical control space. The purpose of the Part 1a project is to work…
Apr 26, 2024
Kobus Esterhuysen

Fly a Drone to a Target using Active Inference

Navigate to a specified target location using Bayesian Inference and RxInfer
Aerospace Industry
Bayesian Inference
Active Inference
RxInfer
Julia
Navigating a drone to a specified target is a commonly encountered problem. In the previous project an active inference agent placed a unit far enough from a radar source to…
Apr 6, 2024
Kobus Esterhuysen

Under the Radar with Active Inference

Finding a safe distance from enemy radar using Bayesian Inference and RxInfer
Military Industry
Bayesian Inference
Active Inference
RxInfer
Julia
Military operating units (for example artillery units, tank units, ships, combat and reconnaissance aircraft) can often be detected by probing enemy radar systems (for…
Apr 3, 2024
Kobus Esterhuysen

District Manager Engagements Analyzed with Active Inference

Using Bayesian Inference and RxInfer to model resource assignments
Retail Industry
Bayesian Inference
Active Inference
Hidden Markov Model
RxInfer
Julia
In this project the client is a newly appointed district manager that is responsible for the optimal running of 4 retail stores in Washington state at the locations:
Mar 20, 2024
Kobus Esterhuysen

A Litter Model using Active Inference

Using Bayesian Inference and RxInfer to estimate daily litter events
Urban Management Industry
Bayesian Inference
Active Inference
RxInfer
Julia
In this project the client is responsible for the delittering of a mile-long beach walk-way in the Pacific Northwest in the USA. The density of foot traffic is roughly…
Mar 15, 2024
Kobus Esterhuysen

Automatic and optimal shift scheduling using Reinforcement Learning (Part 5)

Using Sequential Decision Analytics to find ongoing optimal decisions
Retail Industry
Scheduling
Powell Unified Framework
Reinforcement Learning
Python
In the previous part we found that the agent exploited the situation by having thCumShifts and thSickProb near their max values so that it can use the complete pool. So…
Nov 4, 2023
Kobus Esterhuysen

Automatic and optimal shift scheduling using Reinforcement Learning (Part 4)

Using Sequential Decision Analytics to find ongoing optimal decisions
Retail Industry
Scheduling
Powell Unified Framework
Reinforcement Learning
Python
In the previous part we added tracking of the cumulative unmet resource demands (Ucum) by resource type. We now also track the overall or total unmet demands. We are also…
Nov 2, 2023
Kobus Esterhuysen

Automatic and optimal shift scheduling using Reinforcement Learning (Part 3)

Using Sequential Decision Analytics to find ongoing optimal decisions
Retail Industry
Scheduling
Powell Unified Framework
Reinforcement Learning
Python
In the previous part we added day-of-week availabilities for each resource.
Oct 31, 2023
Kobus Esterhuysen

Automatic and optimal shift scheduling using Reinforcement Learning (Part 2)

Using Sequential Decision Analytics to find ongoing optimal decisions
Retail Industry
Scheduling
Powell Unified Framework
Reinforcement Learning
Python
In the previous part we setup an AI agent to provide automatic and optimal scheduling to assist an HR manager.
Oct 24, 2023
Kobus Esterhuysen

Automatic and optimal shift scheduling using Reinforcement Learning (Part 1)

Using Sequential Decision Analytics to find ongoing optimal decisions
Retail Industry
Scheduling
Powell Unified Framework
Reinforcement Learning
Python
In this project the client had a need to be convinced of the benefits of formal optimized sequential decision making. The client’s need relates to human resource scheduling.…
Oct 21, 2023
Kobus Esterhuysen

Manage Fractional HR Resources with Reinforcement Learning

Using Sequential Decision Analytics to schedule assignments optimally
Consulting Industry
Management Industry
Powell Unified Framework
Reinforcement Learning
Python
In this project the client had a need to be convinced of the benefits of formal optimized sequential decision making. The client’s need relates to automated and optimal…
Oct 18, 2023
Kobus Esterhuysen

Car Rental Operations using the Powell Unified Framework (Part 2)

Using Sequential Decision Analytics to find ongoing optimal decisions
Automotive Industry
Car Rental Industry
Powell Unified Framework
Reinforcement Learning
Python
In this project the client had a need to be convinced of the benefits of formal optimized sequential decision making. The client’s need relates to human resource scheduling.…
Oct 13, 2023
Kobus Esterhuysen

Car Rental Operations using the Powell Unified Framework (Part 1)

Using Sequential Decision Analytics to find ongoing optimal decisions
Automotive Industry
Car Rental Industry
Powell Unified Framework
Reinforcement Learning
Python
In this project the client had a need to be convinced of the benefits of formal optimized sequential decision making. The client’s need relates to human resource scheduling.…
Oct 3, 2023
Kobus Esterhuysen

Dynamic Trading

Using the Powell Unified Framework & Sequential Decision Analytics to find the optimal time to buy and sell assets
Investment Industry
Inventory Management
Powell Unified Framework
Reinforcement Learning
Python
The overall structure of this project and report follows the traditional CRISP-DM format. However, instead of the CRISP-DM’S “4 Modeling” section, we inserted the “6 step…
Aug 8, 2023
Kobus Esterhuysen

Asset Selling

Using the Powell Unified Framework & Sequential Decision Analytics to find the optimal time to sell an asset
Investment Industry
Inventory Management
Powell Unified Framework
Reinforcement Learning
Python
The overall structure of this project and report follows the traditional CRISP-DM format. However, instead of the CRISP-DM’S “4 Modeling” section, we inserted the “6 step…
Jul 11, 2023
Kobus Esterhuysen

Blood Management in a Hospital

Using the Powell Unified Framework & Sequential Decision Analytics to find ongoing optimal decisions
Healthcare Industry
Inventory Management
Powell Unified Framework
Reinforcement Learning
Python
CVXOPT
Linear Programming
The overall structure of this project and report follows the traditional CRISP-DM format. However, instead of the CRISP-DM’S “4 Modeling” section, we inserted the “6 step…
Jun 1, 2023
Kobus Esterhuysen

Demand Simulation of a Consumer Product

Investigating the effect of marketing actions on demand
Marketing Industry
Marketing
Powell Unified Framework
Python
In today’s highly competitive business landscape, effective marketing is crucial for any organization aiming to stand out from the crowd and attract customers. One powerful…
Jan 17, 2023
Kobus Esterhuysen

Learning the best medical treatment

Using Sequential Decision Analytics to find the optimal treatment
Healthcare Industry
Optimal Treatment
Powell Unified Framework
Reinforcement Learning
Multi-Armed Bandits
Python
In this project the client had a need to investigate the use of the NBA (Next Best Action) technique to optimize marketing effort and to have a better handle on customer…
Jan 3, 2023
Kobus Esterhuysen

Dealership Inventory Management using the Powell Unified Framework (Part 5)

Using Sequential Decision Analytics to find ongoing optimal decisions
Automotive Industry
Inventory Management
Powell Unified Framework
Reinforcement Learning
Gymnasium
Python
In part 4 we learned two parameters to provide the optimal partitioning of lot space (in addition to learning the optimal buy levels).
Nov 21, 2022
Kobus Esterhuysen

Dealership Inventory Management using the Powell Unified Framework (Part 4)

Using Sequential Decision Analytics to find ongoing optimal decisions
Automotive Industry
Inventory Management
Powell Unified Framework
Reinforcement Learning
Python
In part 3 we added stockout costs as well as holding costs to complicate Mr. Optimal’s task of managing the number of Elantras and Sonatas on his dealership space of 57…
Nov 7, 2022
Kobus Esterhuysen

Dealership Inventory Management using the Powell Unified Framework (Part 3)

Using Sequential Decision Analytics to find ongoing optimal decisions
Automotive Industry
Inventory Management
Powell Unified Framework
Reinforcement Learning
Python
In part 2 Mr. Optimal managed the number of Elantras and Sonatas on his dealership space of 57 lots. In this project we expand the model so that stockout costs as well as hol…
Oct 14, 2022
Kobus Esterhuysen

Dealership Inventory Management using the Powell Unified Framework (Part 2)

Using Sequential Decision Analytics to find ongoing optimal decisions
Automotive Industry
Inventory Management
Powell Unified Framework
Reinforcement Learning
Python
In part 1 we had Mr. Optimal manage the number of Elantras on his dealership space. In this project we expand his reponsibility to also manage the number of Sonatas.
Sep 28, 2022
Kobus Esterhuysen

Dealership Inventory Management using the Powell Unified Framework (Part 1)

Using Sequential Decision Analytics to find ongoing optimal decisions
Automotive Industry
Inventory Management
Powell Unified Framework
Reinforcement Learning
Python
The overall structure of this project and report follows the traditional CRISP-DM format. However, instead of the CRISP-DM’S “4 Modeling” section, we inserted the “6 step…
Sep 21, 2022
Kobus Esterhuysen

Energy Storage using the Powell Unified Framework (Part 4)

Using Sequential Decision Analytics to find ongoing optimal decisions
Energy Storage
Powell Unified Framework (PUF)
Reinforcement Learning (RL)
In Part 4 we also change the 1-hot encoding of the decision variable to something more scalable. We will have a single decision variable (\(x_{t,GB}\)) where a positive…
Sep 18, 2022
Kobus Esterhuysen

Energy Storage using the Powell Unified Framework (Part 3)

Using Sequential Decision Analytics to find ongoing optimal decisions
Energy Storage
Powell Unified Framework (PUF)
Reinforcement Learning (RL)
In Part 2 we used 1-hot encoding for the decision variable. We also optimized a Bellman policy - one in the Value Function Approximation (VFA) class. This policy was be…
Sep 16, 2022
Kobus Esterhuysen

Energy Storage using the Powell Unified Framework (Part 2)

Using Sequential Decision Analytics to find ongoing optimal decisions
Energy Storage
Powell Unified Framework (PUF)
Reinforcement Learning (RL)
In Part 1 we used 1-hot encoding for the decision variable. We also optimized a buy-low-sell-high policy from the Policy Function Approximation (PFA) class. In addition, we…
Sep 15, 2022
Kobus Esterhuysen

Energy Storage using the Powell Unified Framework (Part 1)

Using Sequential Decision Analytics to find ongoing optimal decisions
Energy Storage
Powell Unified Framework (PUF)
Reinforcement Learning (RL)
The overall structure of this project and report follows the traditional CRISP-DM format. However, instead of the CRISP-DM’S “4 Modeling” section, we inserted the “6 step…
Sep 12, 2022
Kobus Esterhuysen

Next Best Action (NBA) modeling with Reinforcement Learning

Selecting the optimal next action to steer a customer along its journey
Advertizing Industry
Marketing Industry
Customer Journey
Next Best Action
Reinforcement Learning
Powell Unified Framework
Gymnasium
Ray
RLlib
Python
This is a POC that shows how Reinforcement Learning could be used to build a strategic Next Best Action (NBA) model. This kind of model is handy to analyze customer jounrneys…
Aug 11, 2022
Kobus Esterhuysen

An RL Approach for Inventory Management (Part 11)

Tabular Temporal-Difference (Q-Learning) Control
Inventory Management
Markov Decision Process
TD
QLearning
In previous projects we have only dealt with the prediction problem. In this project we move on to the control problem. We stay in the model-free or learning domain (rather…
Jul 22, 2022
Kobus Esterhuysen

An RL Approach for Inventory Management (Part 10)

Tabular Temporal-Difference (SARSA) Control
Inventory Management
Markov Decision Process
TD
SARSA
In previous projects we have only dealt with the prediction problem. In this project we move on to the control problem. We stay in the model-free or learning domain (rather…
Jul 21, 2022
Kobus Esterhuysen

An RL Approach for Inventory Management (Part 9)

Tabular Monte-Carlo Control
Inventory Management
Markov Decision Process
MC
In previous projects we have only dealt with the prediction problem. In this project we move on to the control problem. We stay in the model-free or learning domain (rather…
Jul 20, 2022
Kobus Esterhuysen

An RL Approach for Inventory Management (Part 8)

Tabular Temporal-Difference Prediction
Inventory Management
Markov Decision Process
TD
So far, we always had a model for the environment. This model came in the form of transition probabilities. It is often the case in the real world that we do not have these…
Jul 19, 2022
Kobus Esterhuysen

An RL Approach for Inventory Management (Part 7)

Tabular Monte-Carlo Prediction
Inventory Management
Markov Decision Process
MC
So far, we always had a model for the environment. This model came in the form of transition probabilities. It is often the case in the real world that we do not have these…
Jul 18, 2022
Kobus Esterhuysen

An RL Approach for Inventory Management (Part 6)

Tabular as a Form of Function Approximation
Inventory Management
Markov Decision Process
Python
We now take a step back. Let us again consider a finite set of \(x\)-values \(\mathcal X={x_1,x_2,...,x_n}\). We further assume that all data sets consisting of \((x,y)\) pai…
Jul 14, 2022
Kobus Esterhuysen

An RL Approach for Inventory Management (Part 5)

Introduction to Function Approximation
Inventory Management
Function Approximation
Python
When the state space becomes very large it is not feasible or even possible to make use of a tabular representation for a MDP (Markov Decision Process). Assume we have…
Jul 13, 2022
Kobus Esterhuysen

An RL Approach for Inventory Management (Part 4)

Using a Markov Decision Process for optimal Costs/Rewards in inventory management
Inventory Management
Markov Decision Process
Python
It is important for an inventory manager to maintain inventory at optimal levels for each kind of product. Too much inventory incurs a holding cost while too little…
Jul 10, 2022
Kobus Esterhuysen

An RL Approach for Inventory Management (Part 3)

Simple Inventory Example with Unlimited Capacity (Infinite State/Action Space)
Inventory Management
Markov Decision Process
Python
It is important for an inventory manager to maintain inventory at optimal levels for each kind of product. Too much inventory incurs a holding cost while too little…
Jul 5, 2022
Kobus Esterhuysen

Solve Dynamic Programming Algorithms using Fixed-Point Theory

A numerical technique to solve Finite Markov Decision Processes with large state/action spaces
Dynamic Programming
Fixed Point Theory
Markov Process
There are 3 classical dynamic programming algorithms: - Policy Evaluation algorithm - calculate the Value Function of a Finite MDP evaluated with a fixed policy \(\pi\) (this…
Jul 2, 2022
Kobus Esterhuysen

An RL Approach for Inventory Management (Part 2)

Using a Markov Reward Process for Costs/Rewards in inventory management
Inventory Management
Markov Reward Process
Python
It is important for an inventory manager to maintain inventory at optimal levels for each kind of product. Too much inventory incurs a holding cost while too little…
Jun 26, 2022
Kobus Esterhuysen

An RL Approach for Inventory Management (Part 1)

Using a Markov Process to provide insight into inventory management
Inventory Management
Markov Process
Python
It is important for an inventory manager to maintain inventory at optimal levels for each kind of product. Too much inventory incurs a holding cost while too little…
Jun 24, 2022
Kobus Esterhuysen

Synthetic Commodity Price Data

Using the Markov Property of States to simulate a commodity’s price
Investment Industry
Markov property
Python
In the context of finance we often have to deal with random processes, so-called stochastic processes. A process generates a sequence of random outcomes over time. We will…
Jun 19, 2022
Kobus Esterhuysen

Synthetic Insurance Data

Using the faker package to generate synthetic data
Insurance Industry
Insurance Data Synthesis
faker
It is not always beneficial to make use of pre-existing datasets in the development of data products. To test a system under development, here are some problems that may…
Apr 19, 2022
Kobus Esterhuysen

Customer Churn using Survival Analysis (Part 2)

Find survival probability of a customer using the scikit-survival package
Marketing Industry
Survival Analysis
Customer Churn
scikit-survival
Customer churn is a term that refers to the event that an existing customer (or client) of a company decides to separate from the company. This could mean that the customer…
Mar 16, 2022
Kobus Esterhuysen

Dynamic Portfolio Management with Multi-Armed Bandits

Using Ray RLlib to enable dynamic asset allocation for a portfolio
Investment Industry
Reinforcement Learning
Multi-Armed Bandits
Ray RLlib
OpenAI Gym
Asset allocation refers to the partitioning of available funds in a portfolio among various investment products. Investment products may be categorized in multiple ways. One…
Dec 7, 2021
Kobus Esterhuysen

Dynamic Portfolio Management with Deep Reinforcement Learning (Portfolio of Fidelity® Funds)

DRL is used to provide dynamic asset allocation for a portfolio
Investment Industry
Reinforcement Learning
A2C
PPO
OpenAI Gym
finrl
stable baselines3
pyfolio
Asset allocation refers to the partitioning of available funds in a portfolio among various investment products. Investment products may be categorized in multiple ways. One…
Nov 19, 2021
Kobus Esterhuysen

Dynamic Portfolio Management with Deep Reinforcement Learning

DRL is used to provide dynamic asset allocation for a portfolio
Investment Industry
Reinforcement Learning
A2C
PPO
OpenAI Gym
finrl
stable baselines3
pyfolio
Asset allocation refers to the partitioning of available funds in a portfolio among various investment products. Investment products may be categorized in multiple ways. One…
Nov 11, 2021
Kobus Esterhuysen

Predict Contribution Effort (Part 1)

Estimation of effort to proofread pre-translated sentences
Translation Industry
Regression
AWS-Sagemaker
Spark
EMR
koalas
Machine-translated content in longer documents can usually not be presented to users directly. There is a need for a human proofreader to verify the quality first. Many…
May 19, 2021
Kobus Esterhuysen

Managing Investor Decisions with Reinforcement Learning

RL is used to provide optimal investment decisions for a typical home owner
Investment Industry
Reinforcement Learning
DDPG
OpenAI Gym
TensorFlow
TF-Agents
This project’s purpose is to answer the need of a client who wanted a ‘feel’ for the factors involved in long-term investment for a typical home owner (i.e. not for a…
Feb 3, 2021
Kobus Esterhuysen

Luganda Translation Engine

Build a translation engine (English to another language) from scratch and train with opennmt
Translation Industry
Neural Machine Translation
PyTorch
opennmt
Python
pandas
statsmodels
plotly
seaborn
matplotlib
java
Machine translation is used increasingly to lighten the load of human translators. The critical component here is the translation engine which is a model that takes a…
Nov 9, 2020
Kobus Esterhuysen

Time-series Classification (Deep Learning to Classify Flight Profiles)

CNNs are used after converting time-series to images
Airline Industry
Deep Learning
Timeseries
Image Classification
Computer Vision
CNN
RNN
Python
fastai
Matlab
This project suggests an automatic way for post-flight analysts to undertake classification of flight profiles into useful versus non-useful classes. Instead of using the…
Sep 15, 2019
Kobus Esterhuysen

Real-time Capture of Airline Data from OpenSky

Adoption of Automatic Dependent Surveillance – Broadcast (ADS–B) by airlines is leading to the wide availability of flight data in the public domain
Airline Industry
ADS-B
API
Python
urllib
The gradual adoption of Automatic Dependent Surveillance – Broadcast (ADS–B) by airlines is leading to the wide availability of flight data in the public domain. Wikipedia…
Apr 22, 2019
Kobus Esterhuysen

NLP Topic Modeling using LSI, LDA, and HDP

Using packages gensim (for topic modeling), spacy (for text pre-processing), pyLDAvis (for visualization of LDA topic model), and python-igraph (for network analysis)
NLP
Topic Modeling
Network Analysis
Python
gensim
spacy
pyLDAvis
igraph
The purpose of this pilot project is to investigate the usefulness of topic modeling on a large sermon set. This set forms the basis of outreach initiatives for a Canadian…
Apr 6, 2019
Kobus Esterhuysen

Network Analysis on the Bali network

A comprehensive assessment of the capabilities of the Python igraph package
Network Analysis
Python
igraph
The purpose of this project is to perform a network analysis on the Bali terrorist network data set for the purpose of assessing the capabilities of the Python igraph…
Mar 15, 2019
Kobus Esterhuysen

Network Analysis on USA Airports

Community detection algorithms – Cliques, Edge-betweenness, Leading eigenvector, Fast-greedy, Louvain, Walktrap, Label propagation, InfoMAP, Spinglass
Airline Industry
Network Analysis
Community Detection
Visualization
R
The purpose of this project is to gain insights from a network analysis of USA airports. This network consists of passenger flights between airports in the United States…
Mar 14, 2019
Kobus Esterhuysen

Using Visualization to Make an Effective Case for Using a Machine Translation Tool

Sound visualization principles (Tufte) are applied in this project
Translation Industry
Visualization
Python
seaborn
I am assisting a non-profit with a translation system that they use to translate documents into dozens of languages. The system presents a translator with a single sentence…
Feb 12, 2019
Kobus Esterhuysen

Critique of a Data Product (Annual Report)

Communicating about data is a special art …
Translation Industry
Data Product
Visualization
This project involves the critique of the format and contents of the annual report of a non-profit. The annual report is about the operations of an initiative known as…
Nov 7, 2018
Kobus Esterhuysen

Time-series Regression (Deep Learning to Detect Change Points)

CNNs are used after converting time-series to images
Airline Industry
Deep Learning
Timeseries
Image Regression
CNN
RNN
Python
fastai
This project investigates whether it is feasible to use Convolutional Neural Networks (CNNs) to perform time-series regression, rather than more traditional methods. Before…
Jul 3, 2017
Kobus Esterhuysen

Sentiment Analysis

Is the American Public Willing to Change Course with President Trump?
API
NLP
Python
nltk
R
The purpose of this project is to demonstrate how social media data can be used to capture public sentiment.
Apr 15, 2017
Kobus Esterhuysen

Welcome To My Portfolio of Projects

Welcome! This is a quick introduction …
Jan 13, 2017
Kobus Esterhuysen
No matching items