Motorcycle Buyers SAS
Regression analysis and decision tree analysis in SAS Miner to find what are the factors leading to a purchase of a motorcycle by a person.
Domain
SAS, Python
The Goal:
The primary goal of the Motorcycle Buyers Prediction Project was to uncover the key drivers behind consumer decisions to purchase motorcycles, using a robust, data-driven approach. By applying predictive modeling through SAS Enterprise Miner, the aim was to empower marketers and retailers with actionable insights into what factors most influence bike-buying behavior—such as commute distance, car ownership, and occupation.
1
The Challenge:
Understanding consumer behavior in motorcycle purchases is complex, with influences ranging from income and education to lifestyle and geographic location. One of the core challenges was handling messy real-world data—imputing missing values, reclassifying variable levels (e.g., converting commute distance into categorical), and evaluating model performance across multiple algorithms including decision trees, regression models, and neural networks. Balancing model accuracy while avoiding overfitting remained a key focus throughout.
2
The Result
The final model—a neural network with 7 hidden units—delivered the most accurate predictions, achieving an ASE of 0.2157 and a ROC index of 0.716, signaling strong predictive capability. The analysis revealed that factors like shorter commute distances, fewer owned cars, and specific occupations significantly increased the likelihood of a motorcycle purchase. The project not only highlighted the predictive power of advanced modeling techniques but also demonstrated the value of thoughtful variable engineering in driving business insights.
3
