PREDICTING STUDENT
DROPOUT RISK
STUDENT’S NAME: Malakpalli Asmin Sri Sourya
SUBMISSION DATE: 03-10-2025
INTRODUCTI
OBJECTIVE: ON
• Identify which students are most at risk of dropping out.
• Use machine learning methods (K-NN, K-Means) for prediction and
segmentation.
• Provide actionable insights to support interventions.
DATASET OVERVIEW:
• Number of Records: 100 students.
• Number of Features: 7 columns—PassengerID, Age, Gender, Pluss, Survived,
Embarked, TravelingAlone.
TOOLS AND TECHNIQUES:
• Excel, Python, Google Sheets.
• Methods: EDA, K-NN Classification, K-Means Clustering.
EXPLORATORY DATA
ANALYSIS (EDA)
Key Findings:
• Minimum Age: 1 ; Maximum Age: 70
• Most common Gender: Split between Male/Female
• Most frequent class: Pclass 1 and 3 observed often.
• Survived: Both 0 (did not survive) and 1 (survived) distributed through records.
Chart Title
Common 80
Column Min Max Mean
Values 70
60
Age 1 70 ~36 15, 29, 31
50
Pclass 1 3 ~2 1, 3 40
30
Survived 0 1 ~0.5 0, 1 20
10
0
Age Pclass Survived
Min Max
METHODOLOGY
K-NN CLASSIFICATION:
• Standardize dataset.
• Calculate distance (commonly Euclidean) between a target student
and all others.
• Select k-nearest neighbors (suggest k=5).
• Predict dropout (Survived/Not Survived) based on neighbor majority.
K-MEANS CLUSTERING:
• Randomly initialize centroids for clusters.
• Assign students to nearest centroid.
• Recompute centroids by averaging feature values in each cluster.
• Repeat until clusters stabilize.
RESULT
• K-NN Classification Results:
Show sample student prediction—nearest neighbor
outcomes, actual label, computed distances.
• K-Means Clustering Results:
Provide final clusters with tables of member assignments
and visual plots of clusters in feature space.
INSIGHTS AND LEARNINGS
• Students with lower Age, traveling alone, and lower Pclass
were more likely to be predicted as at risk (or not survived).
• Segmentation using clustering shows distinct groups—e.g.,
older, first-class survivors vs younger, third class non-
survivors.
• Data Science and ML approaches helped reveal patterns not
immediately obvious
CHALLENGES AND
RECOMMENDATIONS
CHALLENGES:
• Missing data or outliers (e.g., extreme ages).
• Computational intensity for manual methods.
RECOMMENDATIONS:
• Automate feature engineering.
• Expand dataset for higher accuracy.
• Use cross-validation for model robustness.
CONCLUSION
• Reviewed the approach and analyzed student dropout risk
using EDA, K-NN, and K-Means.
• The broader implication: Data-driven approaches enable
better early identification and support for at-risk students.
• Emphasized the value of data science/AI in education
applications.
REFERENCES
• Tools:
Excel, Python, Google Sheets.
• Dataset:
Titanic Dataset variant.
• Methodologies:
Standard textbook procedures for EDA, K-NN, and K-Means.