K-Means Clustering —assignment1

Topic: Online Learning Platform — Student Engagement Segmentation

---

1. Business Scenario

"EduSmart" is an online learning platform offering video courses and quizzes. The platform has thousands of students, but the EdTech team has no idea how to categorize them.

They want to know:

• Which students are highly engaged?
• Which students are at risk of dropping out (low engagement)?
• Which students fall in between?

There is no existing label for "engagement level" — nobody has manually tagged the students. The team wants to use unsupervised learning (K-Means) to discover natural student groups based on their behavior, so they can:

• Send reminder emails to low-engagement students
• Offer advanced courses to highly engaged students
• Adjust course content based on group patterns

---

2. Dataset

Two behavioral features are tracked for each student:

• HoursWatched → Total hours of video content watched in a month
• QuizzesAttempted → Number of quizzes attempted in a month

Student	HoursWatched	QuizzesAttempted
S1	1	1
S2	2	1
S3	2	2
S4	3	2
S5	1	2
S6	15	10
S7	16	11
S8	17	12
S9	18	13
S10	16	12
S11	7	5
S12	8	5
S13	8	6
S14	9	6
S15	7	4
S16	22	18
S17	23	19
S18	24	20
S19	25	21
S20	23	20
S21	2	0
S22	9	7
S23	17	13
S24	24	19
S25	1	0

Note: The number of natural groups (K) is not given to you. Part of this assignment is figuring it out yourself using data-driven methods — this mirrors real workplace tasks, where nobody hands you the "right" K.

---

3. Assignment Tasks

Part A — Data Preparation

1. Load the dataset into a Pandas DataFrame.
2. Check for missing values and basic statistics (df.describe()).
3. Select the relevant features for clustering.
4. Scale the features using StandardScaler (important since both features should contribute equally).

Part B — Finding the Right K

5. Use the Elbow Method: run K-Means for K = 1 to 10, record inertia_, and plot the WCSS curve.
6. Use the Silhouette Score: run K-Means for K = 2 to 10 and print the silhouette score for each.
7. Based on both methods, choose and justify the best value of K.

Part C — Building the Final Model

8. Train the final KMeans model using your chosen K.
9. Add a Cluster column to the DataFrame.
10. Print the cluster centers (remember to inverse-transform if you scaled the data, so the centers are interpretable in real hours/quizzes).

Part D — Visualization

11. Create a scatter plot of HoursWatched vs QuizzesAttempted, colored by cluster, with cluster centers marked clearly.

Part E — Business Interpretation

12. Label each cluster in plain English (e.g., "Cluster 0 = Low Engagement / At Risk").
13. Predict the cluster for a new student with HoursWatched=10 and QuizzesAttempted=7.
14. Write a short paragraph (5–6 sentences) recommending what EduSmart should do for each cluster.

---

4. Guiding Questions (answer in your submission)

1. Why did you scale the data before clustering? What might go wrong if you didn't?
2. Did the Elbow Method and Silhouette Score agree on the same K? If not, how did you decide?
3. K-Means assumes clusters are roughly round/spherical in shape. Do you think that assumption fits this dataset? Why or why not?
4. If EduSmart had 10 behavioral features instead of 2, how would that make choosing K and visualizing clusters harder?
5. What is one limitation of K-Means that could cause it to mislabel a student's engagement group?

---

5. Code Skeleton

import pandas as pd
import matplotlib.pyplot as plt
from sklearn.preprocessing import StandardScaler
from sklearn.cluster import KMeans
from sklearn.metrics import silhouette_score

# ---------------------------------
# Part A: Data Preparation
# ---------------------------------
data = {
    "HoursWatched": [1,2,2,3,1,15,16,17,18,16,7,8,8,9,7,22,23,24,25,23,2,9,17,24,1],
    "QuizzesAttempted": [1,1,2,2,2,10,11,12,13,12,5,5,6,6,4,18,19,20,21,20,0,7,13,19,0]
}
df = pd.DataFrame(data)
print(df.describe())

X = df[____]                          # TODO: select features

scaler = StandardScaler()
X_scaled = scaler.____(X)             # TODO: correct method (fit_transform)

# ---------------------------------
# Part B: Finding the Right K
# ---------------------------------
wcss = []
for k in range(1, 11):
    model = KMeans(n_clusters=k, random_state=42, n_init=10)
    model.fit(X_scaled)
    wcss.append(model.____)           # TODO: correct attribute for WCSS

plt.plot(range(1, 11), wcss, marker='o')
plt.xlabel("K")
plt.ylabel("WCSS")
plt.title("Elbow Method")
plt.show()

for k in range(2, 11):
    model = KMeans(n_clusters=k, random_state=42, n_init=10)
    labels = model.____(X_scaled)     # TODO: correct method (fit_predict)
    score = silhouette_score(X_scaled, labels)
    print(f"K={k}, Silhouette Score={score:.3f}")




# ---------------------------------
# Part C: Final Model
# ---------------------------------
final_k = ____                        # TODO: fill in your chosen K
model = KMeans(n_clusters=final_k, random_state=42, n_init=10)
df["Cluster"] = model.fit_predict(X_scaled)

centers_scaled = model.cluster_centers_
centers_original = scaler.____(centers_scaled)   # TODO: correct method (inverse_transform)
centers_df = pd.DataFrame(centers_original, columns=X.columns)
print(centers_df)

# ---------------------------------
# Part D: Visualization
# ---------------------------------
plt.scatter(df["HoursWatched"], df["QuizzesAttempted"], c=df["Cluster"], cmap="viridis")
plt.scatter(centers_df["HoursWatched"], centers_df["QuizzesAttempted"],
            c="red", marker="X", s=200, label="Centers")
plt.xlabel("Hours Watched")
plt.ylabel("Quizzes Attempted")
plt.title("Student Engagement Clusters")
plt.legend()
plt.show()

# ---------------------------------
# Part E: Predict New Student
# ---------------------------------
new_student = [[10, 7]]
new_student_scaled = scaler.____(new_student)     # TODO: correct method (transform, NOT fit_transform)
prediction = model.predict(new_student_scaled)
print("Predicted Cluster:", prediction[0])

---

6. Deliverables

Submit a single notebook/script + short write-up containing:

Item	Description
Code	Fully working K-Means pipeline (Parts A–E)
Elbow plot	Screenshot/plot image
Silhouette scores	Table or printed output for K=2–10
Final cluster scatter plot	With labeled centers
Cluster interpretation	Plain-English label + recommendation per cluster
Guiding question answers	5 short written answers

---

7. Grading Rubric (100 points)

Criteria	Points
Correct data preparation & scaling	15
Elbow Method implemented correctly	15
Silhouette Score implemented correctly	15
Justified choice of K	10
Final model + cluster centers (correctly un-scaled)	15
Visualization (clear, labeled, centers shown)	10
Business interpretation of each cluster	10
Guiding questions answered thoughtfully	10

---

8. Submission Checklist

• [ ] Data loaded and explored
• [ ] Features scaled
• [ ] Elbow method plotted
• [ ] Silhouette scores computed and compared
• [ ] K chosen with justification
• [ ] Final model trained
• [ ] Cluster centers un-scaled and printed
• [ ] Scatter plot with clusters + centers
• [ ] New student prediction made
• [ ] Business interpretation written per cluster
• [ ] All 5 guiding questions answered