How to apply Clustering Techniques to a Cannabis Dataset for Analysis

Musab_Bin_Gulfam · September 2, 2024, 6:31pm

Hi everyone,

This week I finished the Machine Learning Specialization on Coursera offered by deeplearning.ai & Stanford University and I want to start my practise from e-commerece Cannabis dataset I have.

I am currently working with a Cannabis dataset that includes various features related to product strains, effects, and user orders.

I’m interested in applying clustering techniques to this dataset to uncover patterns or groupings that might not be immediately apparent.

Could you please provide guidance on how to effectively employ clustering algorithms, such as K-means in this context? Specifically, how do I choose the right features from the dataset that are most relevent for clustering?

Thanks in advance.

Nevermnd · September 2, 2024, 8:54pm

@Musab_Bin_Gulfam I don’t know if this was covered in the ML course, but you might try PCA to try and rank the features by relevance.

What is less clear though from your description is what is your dependent variable ? What are you predicting ?

ai_curious · September 2, 2024, 9:40pm

Sounds to me like you might benefit from a deep dive into Exploratory Data Analysis on your dataset before you try building unsupervised or supervised ML models.

Here’s an example from the interweb done on the ubiquitous King’s County Housing Data

github.com

teguharia172/King-County-House-Exploratory-Data-Analysis-and-Machine-Learning-Modeling/blob/main/House Sales EDA.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# House Sales in King County, WA\n",
    "\n",
    "Fundamental determinant when it comes to selling houses are the the size of the land and the size house itself.\n",
    "And If you have ever browse house price's through the internet, chances are you know a thing or two about figuring out the value of a property. If the asking price is too high, the likelihood of the home being sold greatly decreases. On the other hand, if a home you are looking to sell is priced too low, you’ll miss out on potential profits.\n",
    "\n",
    "\n",
    "To make sure a home is fairly priced, an analysis towards a house should always be performed, before buying or selling. This analysis will compare the values of homes in King County and what are the factors that affecting price of a house in King County to come up with an accurate market price prediction with machine learning modeling. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Problem Statement\n",

This file has been truncated. show original

Note: I am not endorsing this as the best example. It is one of hundreds out there. Maybe take a read through and steal, eh, reuse ideas that look helpful. Once you have a good idea of the data, picking an ML approach should be more straightforward.

TMosh · September 2, 2024, 11:01pm

Initially, I recommend you include all of the features. You might get useful results without the risk of throwing away features that could be vitally important.