The One-Click Data Scientist: The Power of #GPT-4's New .CSV File Analysis with Python

Having spent the last few weeks exploring #AWS #SageMaker and the seemingly endless models and lines of #Python code, I decided to shut down the #SageMaker virtual instance, conceding because, after all, being a #datascientist is a profession in and of itself with many certifications and specialities right? #Tableau alone has about six certifications, and that’s just a #visualization tool. Well, that fact quickly changed as quickly as it entered my mind.

I’ve experimented with #ChatGPT data analysis before; it's pretty decent when you copy and paste data (non-sensitive) into it. However, I accidentally came across the new Beta option called “#CodeInterpreter” with the ability to upload CSV files! Really! GPT can now accept data files??

So I decided to test it out (using some sample data), and it was extremely impressive. It spent about 3 minutes going step-by-step through a host of data analysis steps - preparation to analysis, determining the most suitable models to identify the right analysis method. It provided detailed findings and graphical outputs.

The steps were:

• Data Loading into a Pandas DataFrame,
• Basic Statistical Analysis (count, mean, standard deviation),
• #EDA to create a histogram illustrating the distribution of variables coupled with a pair plot,
• #Elbow method for #kmeans clustering,
• Additional #kmeans #clustering on selected variables
• Providing comprehensive insights on the analysis.
• About 200 lines of Python code accompany all this.?

The final output:?

• Cluster 1 (Red): Customers with medium annual income and medium spending scores.

• Cluster 2 (Blue): Customers with high annual income but low spending scores. These could be considered "efficient savers".

• Cluster 3 (Green): Customers with high annual income and high spending scores. These could be the "target" group for marketing efforts as they have high spending power and willingness to spend.

• Cluster 4 (Cyan): Customers with low annual income but high spending scores. This group could be risky to target due to lower income.

• Cluster 5 (Magenta): Customers with low annual income and low spending scores.

# Perform K-means clustering with 5 cluster
kmeans = KMeans(n_clusters=5, init='k-means++', random_state=42)
y_kmeans = kmeans.fit_predict(X)

# Plot the clusters
plt.figure(figsize=(10, 7))
plt.scatter(X.iloc[y_kmeans == 0, 0], X.iloc[y_kmeans == 0, 1], s = 100, c = 'red', label = 'Cluster 1')
plt.scatter(X.iloc[y_kmeans == 1, 0], X.iloc[y_kmeans == 1, 1], s = 100, c = 'blue', label = 'Cluster 2')
plt.scatter(X.iloc[y_kmeans == 2, 0], X.iloc[y_kmeans == 2, 1], s = 100, c = 'green', label = 'Cluster 3')
plt.scatter(X.iloc[y_kmeans == 3, 0], X.iloc[y_kmeans == 3, 1], s = 100, c = 'cyan', label = 'Cluster 4')
plt.scatter(X.iloc[y_kmeans == 4, 0], X.iloc[y_kmeans == 4, 1], s = 100, c = 'magenta', label = 'Cluster 5')

# Plot the centroids
plt.scatter(kmeans.cluster_centers_[:, 0], kmeans.cluster_centers_[:, 1], s = 300, c = 'yellow', label = 'Centroids')

plt.title('Clusters of customers')
plt.xlabel('Annual Income (k$)')
plt.ylabel('Spending Score (1-100)')
plt.legend()
plt.show()        

It continued,

Cluster 1 (Red - Medium annual income and medium spending score): This group of customers represents a balanced category. Marketing strategies for this segment should be moderate, focusing on providing value for money and quality products or services.

Cluster 2 (Blue - High annual income and low spending score): This group represents customers with high earning but low spending tendencies - they are more conservative with their spending. These could be potential customers for high-value products or services if marketed effectively. It might be beneficial to investigate why their spending scores are low - are they not finding products they like, or is there another reason for their low spending?

Cluster 3 (Green - High annual income and high spending score): This segment of customers could be the primary target market for the business as they have high income and high spending scores. They are likely to be less price-sensitive and more focused on quality and service. Tailoring high-end products and excellent customer service could be a winning strategy for this segment.

Cluster 4 (Cyan - Low annual income and high spending score): These customers are spending a lot despite having a low income. The company might want to provide budget-friendly options or payment plans for these customers to maintain their loyalty and ensure they don't overextend their finances.

Cluster 5 (Magenta - Low annual income and low spending score): This group of customers might be the most price-sensitive and possibly the most challenging to retain. Offering discounts, lower-priced options, or value deals might be an effective strategy for this group.

For all clusters, it's important to consider other demographic factors like age and gender when crafting marketing strategies. For instance, younger customers might be more responsive to digital marketing campaigns, while older customers prefer more traditional marketing channels.

Finally, these insights are based on the current snapshot of data. Customer behaviour and preferences can change over time, so it's crucial to repeat this analysis periodically to keep the customer segmentation current and relevant.

In conclusion, #gpt4's #python interpreter isn't just an upgrade; it seems like another game-changer. It empowers you to unlock the power of a data scientist with a single click. No longer is data analysis a labyrinthine process reserved for the dedicated few. With #gpt4, it's now a journey we can all embark on.

About the Author: Stuart Bateman is a part-time doctorate candidate at the University of Swansea, UK. His research focuses on AI in Marketing. Stuart is the Head of Digital Marketing at geidea - the leading Fintech in the Middle East, and an Elected Member of the Chartered Institute of Marketing.