Diabetes Diagnosis Based on Blood Tests

🩺 Diabetes Diagnosis Based on Blood Tests

Problem Description

The goal is to build a classification model that predicts whether a patient has diabetes based on blood tests and demographic data.

Each patient is characterized by 8 numerical attributes obtained from analyses and clinical measurements, and the label (target) indicates the presence of diabetes (1 for positive, 0 for negative).

This type of problem belongs to the binary classification category.

---

🔹 Features

• pregnancies – number of pregnancies
• glucose – blood glucose level
• blood_pressure – blood pressure
• skin_thickness – skin fold thickness
• insulin – insulin level
• bmi – body mass index
• diabetes_pedigree_function – genetic risk score
• age – patient age

---

📘 Input File Structure

train.csv

Contains all 8 feature columns plus the column:

• target – indicates the presence of diabetes (0 or 1)

Example:

SampleID	pregnancies	glucose	blood_pressure	skin_thickness	insulin	bmi	diabetes_pedigree_function	age	target
1	6	148	72	35	0	33.6	0.627	50	1
2	1	85	66	29	0	26.6	0.351	31	0

test.csv

Contains the same columns without target, but includes SampleID.

Example:

SampleID	pregnancies	glucose	blood_pressure	skin_thickness	insulin	bmi	diabetes_pedigree_function	age
1	0	137	40	35	168	43.1	2.288	33
2	5	116	74	0	0	25.6	0.201	30

---

📤 Submission

The output file (submission.csv) must contain exactly two columns:

• SampleID
• label – the label predicted by the model (0 or 1)

Example:

SampleID	label
1	1
2	0
3	0

---

⚙️ Evaluation

Model evaluation will be performed using the following metric:

• Binary F1-score

This metric is suitable for binary classification because it gives equal importance to prediction accuracy for both classes.

General formula:

where:

• 
• 

The final score is expressed as a percentage (0–100), rounded to two decimal places.

---

📊 Source

The dataset comes from the original collection:
Pima Indians Diabetes Database – Kaggle