Page 70 - Kaleidoscope Academic Conference Proceedings 2024

2024 ITU Kaleidoscope Academic Conference

advanced ones, are expected to play a significant role in
increasing the accuracy and the level of personalization in
the respective field. This introduction will describe the
importance of prediabetes screening technology and the role
of ML and QML in this new domain. Also, this technology
would be able to change the healthcare management of
diabetes for the better.

An expanded set of algorithms including LR, SVM, Decision
Tree (DT), and Random Forests (RF) has been applied to
diabetes prediction using labeled data that allows to build
the model capable of constructing classes reflecting various
degrees of diabetes risk. They are good at realizing the
interconnections between inputs and outcomes of disease by
which precise risk assessments can be carried out and then
direct interventions are taken. Besides, Ensemble Learning
(EL) approaches that exploit the cooperation of many
models for settlement of superiority have well proven their
considerable role in enhancing the competence and overall Figure 3 – Machine Learning Techniques Classification.
correlation of predictors of diabetes. The Contribution of
discriminating hyperplane between the two classes and
this paper is as follows:
frequently applies advanced kernel strategies that help to deal
• Conducted experiments on the Indian PIMA diabetes with non-linear separability.
dataset to evaluate the algorithms.
Tree-based [9] methods also function based on the result of
• Addressed data imbalance using Synthetic Minority splitting the data based on the feature value and building
Over-sampling Technique (SMOTE) for improved structures easy to comprehend in terms of class labels.
model accuracy. However, these models can be overfitted if the pruning or
even the ensemble methods are not included. Memory-based
• Applied Principal Component Analysis (PCA) for
algorithms [10] classify data by identifying the majority
feature extraction, enhancing computational efficiency
class within the nearest neighbors list. This approach is
and predictive performance.
quite simple, although can be computationally expensive and
• Analyzed the performance of ML and QML algorithms lacks the method of tuning most parameters. Even though
for diabetes prediction. probabilistic models of feature independence are used, these
methods are strikingly accurate for large-sized features like
The rest of the paper consists of the following sections: text classification for instance. The models [11] that mimic
Section 2 discusses machine learning techniques; Section the neuron’s connected structures are particularly efficient
3 includes basics of quantum computing; Section 4 consists in working with vast and multivarious inputs successfully
of various QML techniques; Section 5 discusses performance applying the models to image and voice recognition. Moving
evaluation and implementation; Section 6 consists of Results to a higher level of the artificial intelligence hierarchy, Deep
and discussion and finally conclusion and future work is Learning, a subcategory of Machine Learning, recreates in
discussed in section 7 and 8 respectively. a way the brain’s capability to decipher complex patterns in
data.
2. MACHINE LEARNING TECHNIQUES
These classification techniques are still significant in
Classification is an essential application of machine learning identifying the pathogenic factors and physiological
[4] when it comes to partitioning data into predefined components of diabetes that provide new chances for
categories, and it has numerous uses in almost every screening and diagnosing diabetes [12]. Nevertheless, due
sector including [5]; medical, financial, and promotional to the rather high rate of diabetes incidence, it is still topical
sectors. Thus, in the case of predicting diabetes [6], such or, more accurately, it becomes the subject of research to
methods are useful, along with data mining and artificial a certain extent; this makes it possible to collect large
intelligence. They can accurately and stably identify the amounts of data. Data mining [13] is considered a useful
common characteristics of diabetes from large amounts of method of comprehending and implementing this data: The
data. One way [7] inputs are classified is by computing the techniques apply both objectively descriptive and predictive
posterior probability of an input belonging to the intended plans. Machine learning is also on the rise and can
class based on the linear model. This approach is quite simple perform those jobs that are socially useful such as automatic
and good for applications where the data is equally simple pattern recognition just like humans’ brains. The choice
and organized linearly maybe through a linear separation. of the particular classification method is governed by the
Another technique [8] aims at finding the best possible selected problem type, characteristics of data to be classified,

– 26 –