Page 152 - Emerging Trends and Innovations in Web-Based Applications and Technologies

International Journal of Trend in Scientific Research and Development (IJTSRD) @ www.ijtsrd.com eISSN: 2456-6470
Priyanka et al. [12] carried out chronic kidney disease analysis techniques, we will be able to find promising results
prediction through naive bayes. Tey have tested using other that enhance the quality of patient data and inspect of
algorithms such as KNN (K-Nearest Neighbor Algorithm), specific items that are related to ML algorithms in medical
SVM (Support Vector Machines), Decision tree, and ANN care.
(Artificial Neural Network) and they have got Naïve Bayes The main purpose of our research paper is to make hospital
with better accuracy of 94.6% when compared to other tasks easy and to develop an efficient and feasible software
algorithms.
that replaces the manual prediction system into an
Mohammed and Beshah [13] conducted their research on automated healthcare management system and also it
developing a self-learning knowledge-based system for enables healthcare providers to improve operational
diagnosis and treatment of the first three stages of chronic effectiveness, reduce medical errors and time consumption.
kidney have been conducted using machine learning. A small If disease can be predicted, then early treatment can be given
number of data have been used in this research and they to the patients which can reduce the risk of life and save life
have developed prototype which enables the patient to of patients. The cost to get treatment of diseases can also be
query KBS to see the delivery of advice. Tey used decision reduced up to an extent by early recognition.
tree in order to generate the rules. The overall performance
Our proposed framework aims to predict chronic diseases
of the prototype has been stated as 91% accurate.
such as heart, kidney, cancer, and diabetes using:
Salekin and Stankovic [9] did evaluation of classifiers such as
 A Hybrid Architecture: Combines features from text
K-NN, RF and ANN on a dataset of 400.
data (processed with TF-IDF) and medical images
Wrapper feature selection were implemented and five (processed using ResNet-18).
features were selected for model construction in the study.  Multi-Modal Learning: Merges textual and visual
The highest classification accuracy is 98% by RF and a RMSE modalities to improve prediction accuracy compared to
of 0.11. S. Tekale et al. [10] worked on “Prediction of Chronic single-modal systems..
Kidney Disease Using Machine Learning Algorithm” with a
dataset consists of 400 instances and 14 features. Tey have  Scalability: Designed to handle additional data types,
used decision tree and support vector machine. Te dataset ensuring broad applicability in clinical settings.
has been preprocessed and the number of features has been
reduced from 25 to 14. SVM is stated as a better model with Table 1: Data Preprocessing Steps
an accuracy of 96.75%. Step Text Data Image Data
Feature TF-IDF ResNet-18 Feature
Xiao et al. [11] proposed prediction of chronic kidney disease Extraction Transformation Extraction
progression using logistic regression, Elastic Net, lasso Numeric RGB
regression, ridge regression, support vector machine, Normalization Standardization Normalization
random forest, XGBoost, neural network and k-nearest Label Encoding,
neighbor and compared the models based on their Encoding Binary Encoding N/A
performance. Tey have used 551 patients’ history data with Feature
proteinuria with 18 features and classified the outcome as Dimensions 1200 128
mild, moderate, Debal and Sitote Journal of Big Data (2022)
9:109 Page 3 of 19 severe. Tey have concluded that Logistic IV. PROPOSED RESEARCH MODEL
regression performed better with AUC of 0.873, sensitivity The proposed model integrates both machine learning (ML)
and specificity of 0.83 and 0.82, respectively. and deep learning (DL) techniques to predict chronic
diseases. It adopts a hybrid multi-modal approach,
Almasoud and Ward [13] aimed in their work to test the combining text-based medical data analysis with image-
ability of machine learning algorithms for the prediction of
based diagnostics to enhance prediction accuracy.
chronic kidney disease using subset of features. Tey used
Pearson correlation, ANOVA, and Cramer’s V test to select 1. Text-Based Analysis
predictive features. Tey have done modeling using LR, SVM,  Feature Extraction: Patient symptoms, lifestyle, and
RF, and GB machine learning algorithms. Finally, they medication data are transformed using TF-IDF
concluded that Gradient Boosting has the highest accuracy vectorization and categorical encoding.
with an F-measure of 99.1.  Model Input: Text features (e.g., symptoms,
demographic details) with up to 1200 dimensions.
Most previously conducted researches focused on two
classes, which make treatment recommendations difficult 2. Image-Based Analysis
because the type of treatment to be given is based on the  Feature Extraction: Medical images (X-rays, CT scans)
stages as our project focuses on chronic disease prediction are processed through a pre-trained ResNet-18 model
using machine learning models based on the dataset with big for feature extraction (128 dimensions).
size and recent than online available dataset
 Preprocessing Steps: RGB conversion, resolution
III. PROPOSED WORK standardization, normalization, and augmentation for
Due to the low-progress nature of Chronic Diseases, it is improved model robustness.
important to make an early prediction and provide effective 3. Hybrid Integration
medication. Therefore, it is essential to propose a decision  Text and image features are combined into a unified
model which can help to diagnose chronic diseases and vector of 1328 dimensions.
predict future patient outcomes. While there are many ways
 A neural network processes this combined feature
to approach this in the field of AI, the present study focuses
vector using two hidden layers with dropout for
distinctly on ML predictive models used in the diagnosis of
regularization, culminating in disease classification.
Chronic Diseases. In comparison to the conventional data
IJTSRD | Special Issue on Emerging Trends and Innovations in Web-Based Applications and Technologies Page 142