Capstone Projects

Classifying Pumpkin Variety by Seed Morphological Characteristics

Program: Data Science Master's Degree
Location: Not Specified (remote)
Student: Eldon Komppa

This report set out to answer the question of whether an algorithm could be used to identify pumpkin seeds to a high enough degree of accuracy to replace seed analysts. This study used the approach of casting a wide net. Many of the most common statistical algorithms seen in prior academic reports regarding seeds were used and were vigorously tweaked. 

The objectives of this study are the following: 

  1. Using any model, achieve an accuracy rate higher than 90%. 
  2. Discover which analytical methods performed best according to accuracy. 
  3. Identify the most and least important morphological features for predictions. 
  4. Discuss the algorithm’s benefits in a business application. 
  5. Recording results and providing their interpretation.