July 15, 2024

Darcy Hagey

Digital Breakthroughs Progress

Hoarder Smashing via Unsupervised Learning

4 min read
Hoarder Smashing via Unsupervised Learning
Hoarder Smashing via Unsupervised Learning


I’m going to lay out a simple problem, and then propose a solution. I’ll explain how we built the solution, and then show you some results. Finally, I’ll leave on an inspirational note about the future of AI.

What is Unsupervised Learning?

Unsupervised Learning

Unsupervised learning is a type of machine learning that does not rely on labeled data. Instead, it finds hidden patterns in data by identifying clusters and relationships between variables. For example, you might use unsupervised learning to cluster customers based on their spending habits or to identify groups of similar products that share similar attributes.

Introduction to Hoarding

Hoarding is a mental disorder characterized by the excessive collection of items that have little or no value and the inability to properly maintain them. The person suffering from hoarding will often accumulate large amounts of clutter in their home, which causes severe distress and impairment in their day-to-day functioning.

Hoarding can affect anyone, but it is most often seen among older adults who live alone or have family members who live with them. It can also occur as part of another psychiatric disorder such as obsessive-compulsive disorder (OCD), depression or schizophrenia.[1]

Problem Statement

The problem of hoarding is a complex one. For starters, we don’t really know how many people are affected by it–some estimates put the number in the millions. It’s also difficult to determine what constitutes hoarding because there’s no official definition of “too much stuff” or “enough space.” And then there are all those other factors: Is your apartment messy? Are you disorganized? Or do you just have an affinity for collecting things?

If we were going to solve this problem using machine learning techniques (which we will!), these questions would be important because they would allow us to train our algorithm with data from different types of hoarders. But first let’s define what hoarding is:

  • Hoarders collect items that others might consider worthless or useless; for example, stacks of newspapers or boxes full of old receipts.* They may have difficulty throwing away items due to sentimental attachment, financial reasons (e.g., paying taxes), etc.* The result is usually very cluttered living spaces with little room left over for activities like cooking meals or sleeping comfortably.* Hoarders often experience mental health issues such as anxiety disorders or depression which make it harder for them than most people would find acceptable just getting rid those things once in awhile so they can function normally again without feeling overwhelmed by clutter every day when they wake up in their own homes!

Proposed Solution

Our proposed solution to this problem is to use unsupervised learning to learn the characteristics of people who hoard. Unsupervised learning is a type of machine learning in which we do not specify the output ahead of time–the system learns on its own from the data it’s given, instead of being told how to behave. For example, when you take an image and run it through an image classifier like OpenCV (an open-source computer vision library), it gives you back labels for each pixel indicating whether that pixel belongs to an object or background region–but in this case there was no human intervention involved in deciding what those labels should be!

This makes sense: after all, if we want our computers (or robots) to be able to do something useful without us telling them exactly how every step should be done beforehand then there must be some way for them figure out those steps themselves without being explicitly programmed every step along the way…and this is where unsupervised learning comes into play!

Experiments and Results

In order to answer these questions, I trained a model on the dataset and evaluated its performance.

In total, there were 1045 participants in this dataset: 655 were hoarders and 390 were non-hoarders. The distribution of these participants across each category is shown below:

  • H = Hoarder (655)
  • N = Non-Hoarder (390)

In this article, we will use unsupervised learning to learn the characteristics of people who hoard.

In this article, we will use unsupervised learning to learn the characteristics of people who hoard.

Unsupervised learning is a machine learning task where the algorithm learns from data without being given any labels or directions from humans. This is in contrast with supervised learning, where you have training data containing both inputs and expected outputs (or labels). For instance, if you wanted a program that could detect cats in images on its own without needing any training data or labels for what it should look like when there’s a cat present versus when there isn’t one–that would be an example of unsupervised learning!

Hoarding occurs when someone collects items that they do not need and cannot use; they keep these items despite having no value or usefulness in their lives. Hoarders often live in unhygienic conditions due to their excessive clutter; this can lead to health problems such as respiratory issues due to mold growth caused by moisture build-up inside piles of trash bags full of old newspapers lying around everywhere!


In this article, we will use unsupervised learning to learn the characteristics of people who hoard. We will use a dataset from the University of California San Diego (UCSD) called “Hoarding and Cluttering,” which contains information about people who have been diagnosed with hoarding disorder.