CASE STUDY ยท PORTFOLIO PRESENTATION
iLUNGS
System Features
A web-based, machine-learning-enabled lung health monitoring and referral platform built for the residents and public health officials of Floridablanca, Pampanga.
01
Symptom Checker
02
Care Navigation
03
Public Health Analytics
04
Machine Learning Engine
MODULE 01
Symptom Checker
A guided, three-step intake, Patient, Symptoms, Conditions, that turns what a resident is feeling into a ranked list of possible respiratory conditions, backed by an ensemble of three machine learning models.
97.9%
Ensemble model accuracy on unseen test data
iLUNGS
02
ilungs.app / symptoms-checker
01 SYMPTOM CHECKER
Setting expectations before anything else
Before a single symptom is entered, iLUNGS stops the user at a clear disclaimer: this tool assists with self-triage, it does not replace a doctor. Anyone who might panic at the result is pointed straight to a medical facility instead of the checker.
Plain-language medical disclaimer
Required consent checkbox before proceeding
Step 1 of a 3-part guided intake: Patient, Symptoms, Conditions
iLUNGS
03
ilungs.app / symptoms-checker
01 SYMPTOM CHECKER
Privacy, spelled out in full
The same care extends to data handling. A dedicated Privacy Policy explains exactly what information iLUNGS collects and why, before the resident commits to using the tool, matching the trust a health platform has to earn.
Full data-collection disclosure
Consistent consent pattern across the flow
2 for 2 Disclaimer and privacy policy both require explicit consent
iLUNGS
04
ilungs.app / symptoms-checker
01 SYMPTOM CHECKER, STEP 1 OF 3
Patient: a quick lifestyle snapshot
The first step asks four yes/no lifestyle questions, exercise, posture, diet, smoking. These answers feed the same dataset the analytics dashboard later uses to correlate lifestyle with disease, so every checker session doubles as public-health research.
4-question lifestyle intake
Feeds municipal lifestyle analytics
4 Lifestyle factors tracked: exercise, posture, diet, smoking
iLUNGS
05
ilungs.app / symptoms-checker
01 SYMPTOM CHECKER, STEP 2 OF 3
Symptoms: search, or point at the body
Residents search symptoms by name or tap directly on an interactive body diagram. Selected symptoms collect into a running list they can edit or clear, keeping the experience approachable for users with no medical vocabulary.
Searchable symptom picker
Interactive body-map selection
Step 2 Symptoms feed directly into the ensemble ML model
iLUNGS
06
ilungs.app / symptoms-checker
01 SAFETY NET
Catching severe symptoms in real time
The moment a resident logs a red-flag symptom, like coughing with blood, the system interrupts with a direct recommendation to seek in-person care, and offers a one-tap route into the Hospital Finder. Self-triage never delays urgent care.
Real-time severe-symptom detection
One-tap handoff to Hospital Finder
Instant Alert triggers the moment a red-flag symptom is logged
iLUNGS
07
ilungs.app / symptoms-checker
01 SYMPTOM CHECKER, STEP 3 OF 3
Possible conditions, ranked and explained
Results are presented as a ranked list, each scored by match strength rather than dressed up as a diagnosis. Selecting a condition opens an overview, its likely specialist, and background, the same information a resident would want before booking a consultation.
Match-strength scoring per condition
Specialist recommendation included
Step 3 Ends in a specialist recommendation, not a diagnosis
iLUNGS
08
01 TRANSPARENCY
Explaining the score, not just showing it
A dedicated legend breaks match strength into four honest bands, Low, Fair, Moderate, Strong, and repeats the same reminder: this ranks similarity to reported symptoms, not the likelihood of actually having the condition.
4-tier match strength legend
Reinforces the tool's own limits
iLUNGS
09
MODULE 02
Care Navigation
Once a resident knows what they might be facing, iLUNGS gets them to care: a live map of nearby hospitals and clinics, filterable by barangay, plus a full profile of each facility's providers and services.
56+
Facilities mapped across Floridablanca, Pampanga
iLUNGS
10
ilungs.app / hospital-finder
02 HOSPITAL FINDER
Every nearby facility, on one map
The Hospital Finder lists every hospital and clinic within Floridablanca, filterable by barangay, next to a live map that plots each facility as a marker. Users can compare distance at a glance instead of guessing which facility is actually closest.
Filter by barangay
Live geographic map with facility markers
56+ Hospitals and clinics mapped across the municipality
iLUNGS
11
ilungs.app / hospital-finder / overview
02 HOSPITAL OVERVIEW
A full profile before you commit
Tapping a facility opens its overview, address, contact number, hours, and a plain-language summary of the hospital, so residents can decide whether it's worth the trip before they leave home.
Contact details & operating hours
Descriptive overview for informed decisions
3 tabs Overview, Healthcare Providers, and Facilities in one profile
iLUNGS
12
ilungs.app / hospital-finder / facilities
02 HOSPITAL OVERVIEW
Facilities, shown not just listed
A photo grid of each hospital's actual rooms, emergency room, operating room, patient wards, laboratories, replaces the guesswork of a text-only listing, letting residents judge a facility's readiness for themselves.
Visual facility gallery
Emergency, surgical & ward coverage at a glance
6+ Room types shown per facility, from ER to laboratories
iLUNGS
13
ilungs.app / hospital-finder / providers
02 HOSPITAL OVERVIEW
Meet the people providing care
The Healthcare Providers tab lists each facility's doctors by specialization and schedule, so a resident referred to a pulmonologist can confirm one is actually on staff, and when, before making the trip.
Doctors by specialization
Weekday & weekend schedules
By role Providers grouped by specialization, not just by name
iLUNGS
14
MODULE 03
Public Health Analytics
Every symptom check becomes a data point. Health officials get demographic, symptom, condition, hotspot, and lifestyle reports built to target outbreak response and community health programs where they're needed most.
6
Live analytics dashboards for municipal health officials
iLUNGS
15
ilungs.app / admin / analytics / demographics
03 DEMOGRAPHICS
Who's actually being reached
The Overview dashboard breaks down respondents by barangay, age, and gender, giving officials an immediate read on which communities are engaging with the platform, and which ones the next outreach push needs to target.
Respondents per barangay
Age & gender distribution
80 Respondents recorded in this municipality-wide sample
iLUNGS
16
ilungs.app / admin / analytics / symptoms
03 SYMPTOMS
The most common complaints, ranked
Common Symptoms aggregates every checker session into one ranked chart, headache and cough led the data set, filterable by barangay and by month, turning individual check-ins into a live community health signal.
Ranked by frequency
Filter by barangay and date
181 Total symptom reports behind this chart, headache the most common
iLUNGS
17
ilungs.app / admin / analytics / conditions
03 CONDITIONS
Matched conditions across the municipality
The Conditions dashboard rolls up every matched result from the Symptom Checker, surfacing which diseases are trending locally, Strep Throat and Hypersensitivity Pneumonitis topped the sample data, so officials can plan targeted screening.
Most-matched diseases by volume
Same barangay & date filters
71 Respondents behind this data, 36 matched to Strep Throat
iLUNGS
18
ilungs.app / admin / analytics / hotspots
03 HOTSPOTS
Finding where a symptom is clustering
Symptoms Hotspot flags which barangays report a chosen symptom most often, Dampe led for 'cough with blood' in this data set, helping health officials spot a localized cluster before it becomes an outbreak.
Barangay ranking per symptom
Built for early outbreak response
Dampe Top barangay for 'cough with blood' in this data set
iLUNGS
19
ilungs.app / admin / analytics / hotspots
03 HOTSPOTS
...and where a diagnosis is clustering
The same page answers the mirror question for matched conditions: Palmayo led for Pneumonia in this data set. Together, the two hotspot views let officials cross-reference symptom clusters against confirmed-pattern diseases.
Barangay ranking per condition
Filterable by disease and date
Palmayo Top barangay for Pneumonia, 8 of 14 respondents
iLUNGS
20
ilungs.app / admin / analytics / lifestyles
03 LIFESTYLE
The habits behind the diagnosis
Lifestyle data collected in the Symptom Checker's first step comes back here as a municipality-wide picture, in this sample, smoking was the most common lifestyle factor recorded, ahead of diet, posture, and exercise habits.
Aggregated from every checker session
Direct input for community health programs
184 Respondents tracked; smoking the most common lifestyle factor
iLUNGS
21
ilungs.app / admin / analytics / datasets
03 DATASETS
The data the model actually learns from
A Datasets Overview keeps the training data itself transparent, how many records exist per disease, and a Variable-to-Disease Relationship view that shows how strongly a single symptom, like a cough, correlates with a specific condition.
Per-disease dataset counts
Variable-to-disease correlation viewer
97 Datasets on record across 9 tracked respiratory conditions
iLUNGS
22
MODULE 04
Machine Learning Engine
Under the hood: three classifiers, Support Vector Machine, Random Forest, and Logistic Regression, combined into one ensemble model and continuously monitored, so the Symptom Checker's results stay accurate as new data comes in.
94%
Average test-data accuracy across all three base models
iLUNGS
23
ilungs.app / admin / machine-learning
04 TRAINING PIPELINE
247 records, split and scored honestly
iLUNGS trains on 247 labeled records, held at an 80/20 train-test split. Every model is scored on data it has never seen, Random Forest hit 100% on training data but is judged on its 94% test score, the number that actually predicts real-world performance.
80% train / 20% test split
Per-model train vs. test accuracy shown
247 Labeled records used to train and test the models
iLUNGS
24
ilungs.app / admin / machine-learning
04 VALIDATION
Cross-validated, not just tested once
Beyond a single train/test split, each model is scored again with cross-validation, Random Forest at 93.85%, Logistic Regression at 93.38%, SVM at 90.55%, confirming the accuracy holds up across different slices of the data, not just one lucky split.
K-fold cross-validation per model
Guards against overfitting
93.85% Random Forest's cross-validated score, the strongest single model
iLUNGS
25
ilungs.app / admin / machine-learning
04 ENSEMBLE RESULT
Three models, one stronger answer
Combining all three classifiers into an ensemble lifts overall accuracy to 97.98%, a 7.89-point jump over the best single model's test score. Averaging predictions across independent models is what makes the Symptom Checker's results dependable enough to act on.
Ensemble of SVM, Random Forest & Logistic Regression
Averages predictions across all three
97.98% Combined-model accuracy, +7.89 pts over the best single model
iLUNGS
26
THANK YOU
From symptom to specialist, in one platform.
iLUNGS pairs a machine-learning symptom checker with hospital and doctor navigation and municipal-level analytics, built to help residents in Floridablanca, Pampanga catch lung disease earlier and respond to it faster.
97.98%
Ensemble model accuracy
3.76 / 4
Usability rating from respondent survey
56+
Hospitals & clinics mapped
4
Core modules: checker, navigation, analytics, ML
27