npcrites / california-school-analysis

graduation rate / test score prediction for California public high schools

california-school-analysis

Current Roadmap (To be continued)

Overview of Data

• The California Department of Education offers data on the following categories:
  • Absenteeism
  • Accountability
  • Annual Enrollment
  • Assessment
  • CALPADS Unduplicated Pupil Count (UPC)
  • CBEDS School/District Information
  • Discipline
  • District of Choice
  • Graduate and Dropout
  • Post-Secondary Enrollment
  • Private Schools Data
  • Public School and District
  • Stability Rate Data
  • Staff and Course Enrollment
• We plan to focus on stricly high school data.

Creating the dataset/Preprocessing

• n by m array where n = the number of schools and m = number of school features. The features will be drawn from each dataset and joined by school.
• Normalize and scale all data

Dimensionality Reduction

• Principal Component Analysis

Unsupervised Clustering

• Cluster schools using K-Means (or the like)
• Analyze why certain schools where clustered together
• Determine optimal number of clusters to use

Supervised Learning of School Performance

• Design custom objective function that measures student performance/school performance/etc. ???
• Use a supervised model on each cluster, using school performance metric as labels.
• Feature importance analysis: What caused rises in school performance increases/decreases by cluster?