UnSupervised Machine Learning Models

Project Overview- K Means Clustering

Project Title: Segmenting Users with K-Means Clustering on Income vs Spending Score

Problem Statement:

The goal of this project is to segment customers based on their income and spending score using K-Means Clustering. Customer segmentation helps businesses tailor their marketing strategies to different customer groups, enhancing customer satisfaction and increasing revenue.

Dataset Used:

The dataset used for this project contains the following features:

CustomerID: Unique identifier for each customer.

Gender: Gender of the customer.

Age: Age of the customer.

Income: Annual income of the customer.

SpendingScore: A score assigned to the customer based on their spending behavior.

Steps Taken:

1. Collect the Data: Import and preview the dataset. Loaded the dataset using pandas.read_csv() to inspect the first few rows.

2. Data Preprocessing: Prepare the data for clustering. Selected relevant features (Income and SpendingScore) for clustering. Standardized the data  to ensure all features contribute equally to the distance calculations.

3. Create the clusters: Apply K-Means Clustering algorithm. Used the KMeans class from sklearn.cluster to perform clustering.

4. Evaluate and Visualize Clusters: Analyze the clusters formed by visualizing the centroids and cluster centres.

5. Choose right number of clusters: Determined the optimal number of clusters using the Elbow method or The Average Silhouette Score or The Calinski Harabasz Score

6. Analyze and interpret the clusters: Visualized the clustered data to understand the distribution of customers across different segments. Derive actionable insights from the clustering results.

Tools Used- Python, Jupyter Notebook, Pandas, Scikit-learn, Machine Learning, UnSupervised Learning, Segmentation

Project 2: Association Analysis using Apriori Algorithm

Problem Statement:

This is a very interesting business Case Study problem statement for Data Scientists. Product Affnity is one of the problem statement where we have to identify products that are more likely to be purchased together.

Dataset Used:

The dataset provided comes from a leading fashion retailer in Australia. Company A conducted a display advertising campaign for this brand, displaying ads to users and encouraging them to make purchases on the brand’s website. This dataset comprises the sales data for all users who made an online purchase during the first half of October 2017.

Tasks:

1. Provide an overview of the brand’s sales by the following attributes:

Overall: Total sales and revenue within the given time period.

Basket: Average unique quantity and revenue per order.

Attributes:

Time of Day: Analysis of sales based on different times of the day.

Day of Week: Analysis of sales based on different days of the week.

Geography: Sales distribution across different geographic locations.

Payment Type: Analysis of sales based on different payment methods.

Frequency: Number of single vs. multiple purchasers. Frequency of multi-purchases. Identification of any typical attributes of frequent purchasers.

Product Affinity: Identification of products that are more likely to be purchased together.

2. Statistical Segmentation of the Brand’s Audience Based on Revenue : Based on analysis, created segments of the brand’s audience into distinct groups according to revenue.

Defined the segments with clear definitions and characteristics of each segment. Also, provided recommendations on how marketing team can target each of these groups.

Steps Taken:

1. EDA (Data Processing/ cleaning/ filling missing values where required/outlier handling/

2. Feature engineering- created multiple new variables from existing dataset that are relevant and will be useful for applying apriori as well as for visualizations. .

3. Created a master and cleaned dataset, which has all the features/columns needed.

4. Created summaries and insights 

5. Create segments and their personas (users whoo are early bird shoppers, users who do multi txns, users whoo apply coupons etc etc)

6. Using the Apriori data mining algorithm, identify product affinity.