Enhanced Multiple Disease Prediction System

Dr. Ponmagal R.S.

Harsh Deep Devesh Yadav
Associate Professor
Student Student
Depratment of CTECH
Department of CTECH Department of CTECH
Chennai, Tamil Nadu
Chennai, Tamil Nadu Chennai, Tamil Nadu
[email protected]
[email protected] [email protected]

Abstract— The early detection of chronic diseases is critical for

The aim of this research is to design and implement a
improving patient outcomes and reducing healthcare costs.
This research presents a comprehensive multiple disease multiple disease prediction system that utilizes machine
prediction system that focuses on five prevalent health learning algorithms to assess the likelihood of an individual
conditions: diabetes, heart disease, chronic kidney disease, developing diabetes, heart disease, chronic kidney disease,
Parkinson's disease, and Alzheimer's disease. Utilizing various Parkinson's disease, or Alzheimer's disease. This system will
machine learning algorithms, including Support Vector integrate various data preprocessing techniques to enhance the
Machines (SVM) for diabetes and Parkinson's disease, quality and reliability of predictions, ultimately aiding
Random Forest for kidney disease, and Logistic Regression for healthcare providers in risk assessment and preventive care.
heart disease, we analyzed datasets to develop robust
predictive models. The datasets were subjected to rigorous Each disease included in this study has distinct risk factors
preprocessing, including handling missing values, feature and symptomatology. For instance, diabetes is characterized
selection, and standardization to enhance model performance. by elevated blood glucose levels, which can lead to
The models were evaluated using accuracy scores, providing complications such as cardiovascular disease and kidney
insights into their predictive capabilities. Our findings indicate failure if left untreated. Heart disease remains one of the
that the developed system can effectively predict the likelihood leading causes of mortality worldwide, with factors such as
of these diseases, thus serving as a valuable tool for clinicians in hypertension, high cholesterol, and obesity contributing
early diagnosis and management. This work emphasizes the significantly to its development. Chronic kidney disease is
importance of integrating machine learning techniques in often a silent condition that can progress to end-stage renal
healthcare to facilitate timely interventions and improve failure, requiring dialysis or transplantation.
overall patient care. Neurodegenerative disorders like Parkinson's disease and
Alzheimer's disease significantly impact quality of life and
Keywords— Multiple Disease Prediction, Machine Learning, independence, with early detection being crucial for effective
Diabetes, Heart Disease, Chronic Kidney Disease, Parkinson's management and support.
Disease, Alzheimer's Disease, Support Vector Machine (SVM),
Random Forest, Logistic Regression, Predictive Modeling, In this study, we will employ various machine learning
Healthcare Analytics, Early Detection, Disease Diagnosis, Data algorithms, including Support Vector Machine (SVM),
Preprocessing Random Forest, and Logistic Regression, to create robust
models for predicting these diseases. The models will be
trained and evaluated on publicly a vailable datasets, ensuring
a comprehensive analysis of each algorithm's performance.
I. INTRODUCTION (HEADING 1) We will also implement thorough data preprocessing steps to
The increasing prevalence of chronic diseases globally handle missing values, standardize data, and eliminate
poses significant challenges to healthcare systems, irrelevant features, thereby enhancing model accuracy.
necessitating innovative solutions for early detection and
This research not only aims to contribute to the existing
effective management. Among these chronic diseases,
body of knowledge in healthcare analytics but also seeks to
diabetes, heart disease, chronic kidney disease (CKD),
provide a practical tool that can be utilized by healthcare
Parkinson's disease, and Alzheimer's disease stand out due to
professionals to facilitate early diagnosis and intervention. By
their debilitating effects on individuals and their substantial
harnessing the power of machine learning, this multiple
economic burden on society. Early diagnosis and timely
disease prediction system has the potential to significantly
intervention are critical for improving patient outcomes and
improve patient outcomes and promote proactice healtcare
reducing healthcare costs, making predictive modeling a
practices.
valuable tool in contemporary healthcare.
Machine learning has emerged as a transformative II. RELATED WORK
approach in the field of medical diagnostics. By leveraging
Keniya, Rinkal, et al. [1] - An Excel sheet was developed from
large datasets and advanced algorithms, machine learning
an open-source dataset, encompassing symptoms for
techniques can identify patterns and relationships that are
approximately 230 diseases, which included over 1,000
often imperceptible to traditional statistical methods. This
unique symptoms. The input data for various machine
capability enables the development of predictive models that
learning algorithms consisted of individual symptoms, age,
can assist healthcare professionals in making informed
decisions about patient care. and gender.

XXX-X-XXXX-XXXX-X/XX/$XX.00 ©20XX IEEE

Arumugam, K., et al. [2] - This study employed Patel, Jaydutt, et al. [14] - This paper compared the
preprocessing techniques on the Cleveland dataset before performance of various machine learning techniques in
utilizing various classifiers to make predictions regarding predicting heart disease, discussing the implications of their
disease presence. findings.
Zou, Quan, et al. [3] - The researchers analyzed data from Revathy, S., et al. [15] - The authors proposed machine
hospital physical examinations in Luzhou, China, applying learning algorithms for predicting Chronic Kidney Disease
machine learning classification methods such as Decision (CKD), evaluating different classifiers and identifying the
Trees, Random Forests, and Neural Networks for diabetes most effective models.
prediction.
Jindal, Harshit, et al. [16] - The researchers examined the use
Wang, Zixian, et al. [4] - This study investigated chronic of KNN, Logistic Regression, and Random Forest classifiers
kidney disease (CKD) using machine learning methodologies for accurately diagnosing heart disease, emphasizing the goal
based on a dataset from the UCI machine learning repository, of improving prediction capabilities.
specifically analyzing 400 individuals and utilizing the A
Priori association technique.
Salhi, Dhai Eddine, et al. [5] - The research team evaluated a III. PROPOSED WORK
dataset featuring Algerian patients, implementing Neural
In this research, we propose the development of a
Networks (NN), K-Nearest Neighbors (KNN), and Support
comprehensive multiple disease prediction system that
Vector Machine (SVM) techniques. Their findings indicated
leverages advanced machine learning techniques to predict the
that the neural network algorithm consistently produced
likelihood of diabetes, heart disease, chronic kidney disease,
accurate results for heart disease prediction.
Parkinson's disease, and Alzheimer's disease in individuals.
Godse, Rudra A., et al. [6] - This work focused on multiple The primary objective of this system is to provide healthcare
disease prediction using various machine learning professionals with a reliable tool for early diagnosis, thereby
algorithms, including Naïve Bayes, KNN, Decision Trees, facilitating timely interventions and improving patient
Random Forests, and SVM. The system also recommended outcomes.
consulting a healthcare professional based on the findings.
Mohit, Indukuri, et al. [7] - A web application was created to
identify diseases like breast cancer, diabetes, and heart
disease through the use of machine learning models,
including Logistic Regression, SVM, and KNN.
Ahmed, Nazin, et al. [8] - The researchers implemented
preprocessing on two distinct datasets, dividing them into
training and testing sets. A range of machine learning
algorithms were utilized to construct prediction models from
the training data, with performance evaluated against
multiple metrics. The most effective model was then
deployed in a web application to predict diabetes.
Shaikh, F. J., and D. S. Rao [9] - This research reviewed
machine learning and deep learning techniques used in cancer
progression modeling, highlighting the significance of data
supervision for accurate predictions.
Rane, Nikita, et al. [10] - This study explored machine
learning approaches for breast cancer classification and
prediction, assessing the effectiveness of several algorithms.
Alanazi, Rayan [11] - The work involved the preprocessing
of datasets, which were then divided into training and testing
subsets. Various machine learning algorithms were utilized Fig. 1. System Architecture of Proposed Work
to train the model, ultimately preparing it for testing and
evaluation. III) System Architecture
Mujumdar, Aishwarya, and Vb Vaidehi [12] - This research The proposed system architecture is designed to
focused on predicting diabetes using different machine systematically process patient data and provide predictive
learning algorithms, highlighting the classification methods insights. As depicted in Figure 1, the architecture consists of
used. several key components:
Ifraz, Gazi Mohammed, et al. [13] - The study trained Data Collection: The system begins with the collection of
multiple models for accurate kidney disease predictions, relevant health data from various sources, including electronic
employing physiological variables and machine learning health records, patient questionnaires, and publicly available
techniques such as Logistic Regression, Decision Tree, and datasets. This data includes patient demographics, clinical
KNN. measurements, and historical hea lth information.
Data Preprocessing: Raw data often conta ins
inconsistencies and missing values, necessitating
preprocessing steps. This phase includes data cleaning linear kernel option also allows for faster training, making
(removing duplicates and handling missing values), data SVM suitable for medical diagnosis applications.
transformation (normalizing and standardizing features), and
feature selection (eliminating irrelevant or redundant Logistic Regression: Logistic Regression is a popular
features). These steps are crucial for enhancing the accuracy statistical method for binary classification. In heart disease
and reliability of the predictive models. prediction, it estimates the probability of an individual having
the condition based on features like age, blood pressure, and
Model Training: After preprocessing, the cleaned dataset cholesterol levels. One major advantage is its interpretability,
is divided into training and testing sets. Various machine as the coefficients reveal the influence of each feature on the
learning algorithms, including Support Vector Machine outcome. This is particularly valuable in healthcare, where
(SVM), Random Forest, and Logistic Regression, are understanding contributing factors is essential. Logistic
employed to train predictive models. The models a re Regression is computationally efficient and works well with
evaluated based on metrics such as accuracy, precision, recall, smaller datasets, making it ideal for initial heart disease
and F1-score to determine their effectiveness in predicting screening.
each disease.
Random Forest: Random Forest is an ensemble learning
Model Evaluation: The trained models are rigorously method that builds multiple decision trees and merges their
evaluated using the testing dataset. This evaluation not only outputs for accurate predictions. We chose it for chronic
assesses the performance of each model but also helps in kidney disease prediction due to its robustness against
selecting the most suitable algorithm for deployment. Cross- overfitting and capability to handle large datasets with both
validation techniques are applied to ensure that the models numerical and categorical features. By aggregating
generalize well to unseen data. predictions, it reduces variance and enhances accuracy.
Prediction and Output: The final component of the Random Forest also highlights feature importance, helping
system is the prediction module, where healthcare practitioners identify key factors influencing kidney disease
professionals can input patient data to receive predictive risk, making it suitable for complex medical datasets.
insights regarding the risk of developing one or more of the
Decision Trees: Decision Trees are a widely used
targeted diseases. The system will generate a user-friendly
algorithm that classifies data points based on feature va lues
report summarizing the predictions, highlighting risk factors,
through a tree-like model of decisions. The algorithm
and suggesting recommendations for further medical
recursively splits the dataset, creating branches that represent
consultation or lifestyle changes.
different decision paths, resulting in an interpretable model.
User Interface: A simple and intuitive user interface (UI) This is particularly valuable in healthcare, where
will be developed to facilitate easy interaction with the understanding predictions is crucial. Each leaf node
prediction system. Healthcare providers can easily navigate corresponds to a classification outcome, with paths from the
through the system, input patient data, and view results root to leaves representing decision rules. The interpretability
without requiring advanced technical skills. of Decision Trees fosters trust and facilitates better decision-
making.
K-Nearest Neighbour (KNN): K-Nearest Neighbors
III) Future Enhancements (KNN) is a straightforward and effective machine learning
algorithm for classification and regression. It classifies new
Future enhancements to the proposed system may include
data points based on the majority class among their 'k' nearest
the integration of real-time health monitoring through
neighbors in feature spa ce, using a distance metric like
wearable devices, allowing for continuous data collection and
Euclidean distance. KNN is non-parametric, making no
more dynamic predictions. Additionally, implementing
assumptions about data distribution, which allows flexibility
advanced visualization techniques could provide more
with complex patterns. It works well with smaller datasets,
insightful interpretations of the prediction results,
leveraging similarities for accurate predictions. However,
empowering healthcare professionals to make informed
KNN can be computationally intensive for larger datasets
decisions based on comprehensive data analyses.
since it requires calculating distances to all training samples.
By developing this multiple disease prediction system, we Despite this, its simplicity and effectiveness make it a valuable
aim to bridge the gap between healthcare and technology, tool in predictive analytics.
providing practitioners with a powerful tool to combat the
growing burden of chronic diseases effectively. D) Streamlit
Streamlit is an open-source Python library that simplifies
the creation and deployment of web applications for machine
III) Machine Learning Models learning and data science projects. Designed for data scientists
and machine learning engineers, Streamlit allows users to
Support Vector Machine: Support Vector Machine (SVM)
build interactive web apps quickly and easily, without
is a powerful supervised learning algorithm used for
extensive web development knowledge. By enabling users to
classification tasks. It finds the optimal hyperplane to separate
turn data scripts into shareable web applications in minutes,
data points from different classes in high-dimensional space.
Streamlit is ideal for prototyping and showcasing machine
We applied SVM for predicting diabetes and Parkinson’s
learning models, while its automatic user interface generation
disease, benefiting from its ability to handle non-linear
and real-time interactivity enhance the overall user
relationships through kernel functions. By transforming input
experience.
features into higher dimensions, SVM maximizes the margin
between classes, improving accuracy and robustness. The
 associated with heart health. Its interpretability is a significant
Features of Streamlit: advantage in medical applications, allowing clinicians to
Ease of Use: Streamlit's straightforward API allows users understand the influence of individual features on the
to focus on Python code without traditional web development likelihood of heart disease. The model achieved an accuracy
complexities, making it accessible to non-programmers. of about 82%, reflecting its efficacy in classifying individuals
as having or not having heart disease based on the given
Real-time Interactivity: Users can manipulate input features.
parameters and see immediate results in visualizations, which
is useful for exploratory data analysis. 5) Alzheimer's Disease Prediction: In the case of
Alzheimer's disease prediction, we employed K-Nearest
Customizable Widgets: Streamlit offers pre-built widgets Neighbors (KNN). KNN is a simple yet effective algorithm
like sliders and dropdowns, enhancing user engagement and that classifies instances based on the proximity of their
experience. features to those of the training examples. Its performance is
Seamless Integration: It integrates smoothly with popular particularly strong when the decision boundary is irregular,
libraries such as NumPy, Pandas, Matplotlib, and Plotly, which is common in healthcare datasets. The model reached
allowing easy use of existing data tools. an accuracy of approximately 68%, effectively distinguishing
between individuals with and without Alzheimer's disease.
Deployment Capabilities: Streamlit apps can be easily
deployed to the cloud via services like Streamlit Sharing,
Heroku, or AWS, facilitating quick sharing with stakeholders.

IV. RESULLS
1) Diabetes Prediction: In our diabetes prediction
model, we employed the Support Vector Machine (SVM)
algorithm. SVM is particularly effective for binary
classification problems, as it aims to find the optimal
hyperplane that separates the classes in the fea ture space.
Given the complexities of the diabetes dataset, including
features that represent various health indicators, SVM's ability
to handle high-dimensional spaces made it a suitable choice.
The model achieved an accuracy of approximately 77% on the
test set, demonstrating its effectiveness in distinguishing
between diabetic and non-diabetic individuals.
2) Parkinson’s Disease Detection: For Parkinson’s
disease detection, we utilized the Support Vector Machine Fig. 2. Output for Diabetes Prediction system
(SVM) model again, as it is well-suited for medical diagnosis
tasks involving complex data. The Parkinson's dataset consists
of various acoustic feat ures extracted from voice recordings,
which are indicative of the disease. SVM was selected for its
robustness in handling non-linear relationships and high-
dimensional data, essential for accurately predicting the
presence of Parkinson’s disease. The model achieved an
accuracy of around 87%, indicating a strong capability to
identify individuals with the disease based on voice features.
3) Chronic Kidney Disease (CKD) Prediction: In the
CKD prediction task, we opted for the Random Forest
classifier. Random Forest is an ensemble learning method that
builds multiple decision trees and merges their outputs to
improve predictive accuracy. This model is particularly
beneficial for healthcare datasets like CKD, where the feature
set can be heterogeneous and the relationships among features
may not be strictly linear. The Random Forest model achieved
an accuracy of approximately 94%, showcasing its
effectiveness in identifying patients at risk of chronic kidney
disease.
4) Heart Disease Prediction: For predicting heart disease,
we selected the Logistic Regression model. Logistic
Regression is a widely used statistical method for binary
classification problems, making it an appropriate choice for
our dataset, which contains various clinical parameters
 Fig. 5. Output for Alzheimer’s Disease Prediction system

Fig. 3. Output for Heart Disease Prediction system

Fig. 4. Output for Parkinson’s Disease Prediction system

Fig. 6. Output for Chronic Kidney Disease Prediction system

 V. CONCLUSION Comparatively." International Journal of Advanced
The choice of machine learning models was driven by the Research in Computer and Communication Engineering
nature of each dataset and the specific characteristics of the 8.12 (2019): 50-52.
[7] Mohit, Indukuri, et al. "An Approach to detect
diseases being predicted. SVM was favored for its robustness multiple diseases using machine learning algorithm."
in high-dimensional spaces in the diabetes and Parkinson's Journal of Physics: Conference Series. Vol. 2089. No. 1.
models, while Random Forest excelled in handling the IOP Publishing, 2021.
complexities of the CKD dataset. Logistic Regression [8] Ahmed, Nazin, et al. "Machine learning based
provided valuable insights in heart disease prediction due to diabetes prediction and development of smart web
its interpretability, and KNN was selected for its effectiveness application." International Journal of Cognitive
Computing in Engineering 2 (2021): 229-241.
in handling irregular decision boundaries in Alzheimer's
[9] Shaikh, F. J., and D. S. Rao. "Prediction of cancer
predictions. Overall, each model demonstrated satisfactory disease using machine learning approach." Materials
accuracy in its respective application, contributing to the Today: Proceedings 50 (2022): 40-47.
overarching goal of developing a comprehensive multiple [10] Rane, Nikita, et al. "Breast cancer classification and
disease prediction system. prediction using machine learning." International
Journal of Engineering Research and Technology 9.2
(2020): 576-580.
[11] Alanazi, Rayan. "Identification and prediction of
chronic diseases using machine learning approach."
REFERENCES Journal of Healthcare Engineering 2022.1 (2022):
2826127.
[1] Keniya, Rinkal, et al. "Disease prediction from
various symptoms using machine learning." Available at [12] Mujumdar, Aishwarya, and Vb Vaidehi. "Diabetes
SSRN 3661426 (2020). prediction using machine learning algorithms." Procedia
Computer Science 165 (2019): 292-299.
[2] Arumugam, K., et al. "Multiple disease prediction
using Machine learning algorithms." Materials Today: [13] Ifraz, Gazi Mohammed, et al. "[Retracted]
Proceedings 80 (2023): 3682-3685. Comparative Analysis for Prediction of Kidney Disease
Using Intelligent Machine Learning Methods."
[3] Zou, Quan, et al. "Predicting diabetes mellitus with Computational and Mathematical Methods in Medicine
machine learning techniques." Frontiers in genetics 9 2021.1 (2021): 6141470.
(2018): 515.
[14] Patel, Jaydutt, et al. "Heart disease prediction using
[4] Wang, Zixian, et al. "Machine learning-based machine learning." Proceedings of Second International
prediction system for chronic kidney disease using Conference on Computing, Communications, and Cyber-
associative classification technique." International Security: IC4S 2020. Springer Singapore, 2021.
Journal of engineering & Technology 7.4.36 (2018):
1161. [15] Revathy, S., et al. "Chronic kidney disease
prediction using machine learning models."
[5] Salhi, Dhai Eddine, Abdelkamel Tari, and M-Tahar International Journal of Engineering and Advanced
Kechadi. "Using machine learning for heart disease Technology 9.1 (2019): 6364-6367.
prediction." Advances in Computing Systems and
Applications: Proceedings of the 4th Conference on [16] Jindal, Harshit, et al. "Heart disease prediction using
Computing Systems and Applications. Springer machine learning algorithms." IOP conference series:
International Publishing, 2021. materials science and engineering. Vol. 1022. No. 1.
IOP Publishing, 2021.
[6] Godse, Rudra A., et al. "Multiple Disease Prediction
Using Different Machine Learning Algorithms