From,

Smit Anant Pale

TY Student,
[Link] Computer Science Department,
Roll No: 251067
Date 22/11/2025

To,
Dr. Pallavi Mirajkar,
Coordinator,
[Link] Computer Science,
Maharshi Dayanand College Of Arts, Science and Commerce,

Subject: Formal Request for Final Year Project Proposal Approval

Respected Madam,

I am writing to seek your approval for my final year project titled

“Cognitive: Cognitive Medical Appointment System.” I have
clearly defined the project’s objectives, outlined the methodology, and
explained the anticipated outcomes. In line with department
regulations, I have incorporated all feedback provided. I am confident
that this project meets academic standards and will add meaningful
value. I would greatly appreciate your prompt review and approval.
Thank you for your time and consideration.

Sincerely,
Smit Anant Pale
 Cognitive: Cognitive Medical Appointment System

Introduction & Rationale:

The healthcare industry is rapidly transitioning toward digital-first

solutions, yet one of the most fundamental interactions — booking a
medical appointment — still suffers from inefficiencies, manual
processes, and poor information flow in many clinics and hospitals.
Traditional appointment systems rely heavily on fixed scheduling
rules and human decision-making, which often results in long patient
wait times, uneven doctor workloads, high no-show rates, and
overall reduced clinical efficiency. Patients frequently struggle to
identify the right specialist, estimate the urgency of their condition,
or manage appointment changes without friction. These issues
highlight the need for a more intelligent and adaptive system.

A Cognitive Medical Appointment System introduces artificial

intelligence and data-driven decision-making into this process. By
incorporating cognitive computing methods such as natural language
processing, predictive analytics, and intelligent scheduling
algorithms, the system becomes capable of understanding patient
symptoms, prioritizing urgent cases, predicting appointment no-
shows, and recommending optimal time slots. Instead of functioning
as a simple booking tool, the system actively “thinks,” analyzes
patterns, and supports both patients and healthcare staff in making
informed decisions.

The system aims to bridge critical gaps in traditional healthcare

scheduling:

 Helping patients receive timely care based on the urgency of their

symptoms.
 Reducing administrative burden on hospitals and clinics through
automation.
 Increasing doctor productivity by minimizing idle time and managing high-
risk no-show appointments.
 Enhancing patient satisfaction through smoother navigation, personalized
reminders, and quick access to care.
 Objectives:

1. Build a web/mobile-accessible appointment system with user roles:

patient, doctor, receptionist/admin.

2. Implement cognitive features:

o NLP-based symptom intake (quick triage) to suggest

appointment urgency and specialty.

o Predictive model for no-show probability (based on history,

time of day, reminders, distance, etc.) and dynamic
overbooking suggestions.

o Smart slot suggestion to minimize wait times and maximize

utilization.

3. Automated multi-channel reminders (SMS/Email/Push) with smart

timing based on predicted responsiveness.

4. Real-time rescheduling workflow with minimal friction.

5. Admin dashboard: clinic utilization, patient flow analytics, no-show

patterns, and model performance metrics.

6. Privacy & security: encrypt PII at rest and in transit; role-based

access control (RBAC); audit logging.
Scope:

A web-based platform that ingests multimedia (images, video, audio),

runs ML/DL-based manipulation detection, generates cryptographic
fingerprints & signed metadata, and records verification events
immutably on a blockchain ledger. Storage of large files is off-chain
(IPFS/Swarm). Verification can be centralized or distributed across
verification nodes running different detection models; consensus or
multi-signer attestations improve trust and resistance to single-point
compromise. The system supports RBAC, alerts, analytics, and secure
third-party APIs.

Patient registration/login, appointment booking, cancellations,

rescheduling:
Patients can quickly register, log in, and manage their profiles
securely.
They can book, cancel, or reschedule appointments with real-time
availability updates.

Symptom intake questionnaire with NLP

A short symptom form collects basic health details and free-text
symptoms.
An NLP model classifies urgency and suggests the right medical
specialty.

Predictive no-show model

A machine learning model predicts the likelihood of a patient missing
an appointment.
These predictions help optimize scheduling and reduce last-minute
gaps.
Reminder engine
Automated reminders are sent via SMS, email, or notifications at
configurable times.
High-risk patients receive extra reminders to ensure attendance.

Doctor & admin interfaces, analytics dashboard

Doctors and admins get dashboards to view schedules, patient details,
and system insights.
Analytics show trends like no-show rates, appointment volume, and
doctor utilization.

Privacy measures and RBAC

Data is secured through encryption, safe communication, and minimal
data storage.
RBAC ensures patients, doctors, and admins access only their
permitted features.
 Methodology:

High-level components:

1 Frontend

 Patient web/mobile UI (React / React Native or responsive React).

 Doctor & Admin web portal.

2 Backend

 RESTful API ([Link]/Express or Python Flask/FastAPI) for

authentication, appointments, reminders, analytics.
 Database: PostgreSQL (primary relational store) + Redis
(caching, job queue).

3 Cognitive Layer

 NLP symptom-intake: lightweight transformer or fine-tuned

DistilBERT / spaCy classifier to map free-text symptoms →
triage label (low/medium/high).
 Predictive model: XGBoost / Logistic Regression to estimate no-show
risk using engineered features.
 Scheduling optimizer: heuristics + greedy algorithm that uses predicted
no-show to suggest safe overbooking.

4 Reminder & Notification Engine

 Background job worker (Celery / Bull) to schedule and send

SMS/email/push; message templates adapt based on risk score.

5 Security & Privacy

 JWT-based auth, RBAC, HTTPS, AES encryption for sensitive fields,

audit logging
 6 Deployment

 Containerized (Docker) + deployment to a cloud provider or college

server; CI pipeline for testing.

7 Data

 If no real dataset available, generate realistic synthetic dataset

for training/testing; follow privacy best practices.
Tools & Technologies:

Layer / Component Technology / Tools

Frontend: React (or React Native for mobile)

Backend: [Link] + Express OR Python FastAPI

Database: PostgreSQL, Redis

ML: Python, scikit-learn, XGBoost, Hugging

Face Transformers or spaCy

Queues: Celery (Python) or Bull (Node)

Notifications: Twilio (SMS) / SMTP / Firebase Cloud

Messaging

Deployment: Docker, GitHub Actions/GitLab CI,

DigitalOcean/AWS/GCP (testnet/college
VM)
 Timeline:

Week Activities
Nov 29 – Dec 15 finalize requirements, data sources, architecture diagrams,
literature review.

Dec 16 – Dec 24 auth, user models, appointment CRUD, RBAC.

Dec 25 – Dec 30 patient booking & doctor/admin login.

Jan 1 – Jan 11 data preparation, model training, evaluation; integration API

endpoints.

Jan 12 – Jan 27 integrate predictions into booking flow, overbooking

heuristics.

Jan 28 – Feb 10 integrate SMS/Email/Push and templates.

analytics, logs, visualization.
Feb 11 – Feb 27 encryption, RBAC testing, unit & integration tests.
load tests, usability tweaks.
Feb 28 – March 3 final report, user manual, presentation.
Expected Outcomes:

The project delivers a functional web-based prototype where patients

can register, log in, and manage appointments, supported by cognitive
features such as NLP-driven symptom triage and a no-show
prediction model. The system includes an admin dashboard that
presents key analytics like doctor utilization and no-show trends,
helping clinics operate more efficiently. Comprehensive
documentation—covering technical reports, user manuals, and
deployment steps—ensures the solution can be easily reproduced or
extended. A demo and presentation slide deck showcase the
workflow, intelligent modules, and overall system performance for
academic evaluation.

Small Examples (Listed Separately)

 Example: NLP flags chest pain as urgent and suggests a cardiologist.

 Example: Dashboard highlights that Tuesday has high no-show rates.
 Example: Patient receives a reminder two hours before the appointment.
 References:

 Improving Hospital Outpatient Clinics Appointment Schedules by

Prediction Models
[Link]

 No-Show in Medical Appointments with Machine Learning Techniques:

A Systematic Literature Review

[Link]

 Prediction of hospital no-show appointments through artificial

intelligence algorithms

[Link]

 Machine Learning for Prediction of Clinical Appointment No-Shows

[Link]
[Link]

 Machine Learning Approach for Real-Time Prediction of Last-Minute

Medical Appointment No-shows

[Link]

 Consultation length and no-show prediction for improving appointment

scheduling efficiency at a cardiology clinic

[Link]
 A Solution to Reduce the Impact of Patients’ No-Show Behavior on
Hospital Operating Costs: Artificial Intelligence-Based Appointment
System

[Link]

 Predict, then schedule: Prescriptive analytics approach for ML-enabled

sequential clinical scheduling

[Link]

 A Multi-Head Attention Soft Random Forest for Interpretable Patient

No-Show Prediction

[Link]

 Fairness in TabNet Model for Prediction of Hospital No-Show

[Link]

 Data-Driven Distributionally Robust Appointment Scheduling over

Wasserstein Balls

[Link]

 Real-Time Analytics and AI for Managing No-Show Appointments in

Primary Health Care: Before-and-After Study (2025)

[Link]

 AI-Driven Doctor Scheduling for Efficient Patient Appointments —

(2025, IRJAEH)
[Link]

 ML & Predictive Analytics for Appointment Scheduling: Review of

 Techniques & No-Show Determinants from systematic review of no-
show literature

[Link]

 Real-world deployment evidence showing AI-based no-show prediction

+ scheduling improves attendance & resource utilization (AI-BAS
paper)

[Link]

 Use of interpretable ML models in no-show prediction, dealing with

missing data & external variables (weather, demographics) — from
pediatric no-show study

[Link]

 Combination of no-show prediction + appointment length estimation to

better schedule and reduce waiting/idle times — from consultation
length & no-show prediction study
[Link]
 Integrating Appointment Scheduling, No-Show Prediction &
Overbooking Strategies using ML + Scheduling Rules — from Predict,
then schedule (2022)

[Link]

 Use of ensemble and decision-tree based models (Random Forest,

Gradient Boosting) for no-show prediction — from Machine Learning
for Prediction of Clinical Appointment No-Shows

[Link]
[Link]