International Journal of Scientific Research in Modern Science and Technology

ISSN: 2583 -7605 (Online)

© IJSRMST | Vol. 4 | Issue 2 | February 2025
Available online at: https://ijsrmst.com/
DOI: https://doi.org/10.59828/ijsrmst.v4i2.301

UNVEILING NOVEL PREDICTORS OF

ALZHEIMER’S DISEASE: A FUNCTIONAL AND
BEHAVIORAL-BASED CLUSTERING APPROACH
Adene, Gift1; Igwe, J. S.2; Adannaya U. Gift-Adene3; Obidinma Christian Alozie4;
Iweama William Chukwuebuka5
1
Department of Computer Science, Akanu Ibiam Federal Polytechnic, Unwana, Ebonyi State Nigeria
1
Email: [email protected] | [email protected]
2
Department of Computer Science, Ebonyi State University, Abakaliki, Ebonyi State, Nigeria
2
Email: [email protected]
3
Department of Computer Science, Akanu Ibiam Federal Polytechnic, Unwana, Ebonyi State Nigeria
3
Email: [email protected]
4
Clifford University, Owerrinta, Nigeria
4
Email: [email protected]
5
Department of Computer Science, Akanu Ibiam Federal Polytechnic, Unwana, Ebonyi State Nigeria
5
Email: [email protected]

ABSTRACT
Alzheimer’s disease (AD) is a progressive neurodegenerative disorder traditionally diagnosed using
cognitive assessments, which may overlook critical functional and behavioral symptoms. This study
employs statistical and machine learning techniques to analyze a publicly available Kaggle dataset of
Alzheimer’s patients, identifying novel predictors of disease progression. Ethical considerations were
addressed by adhering to secondary data analysis guidelines, with feature selection performed using
correlation analysis and principal component analysis (PCA). Model validation, including 10-fold cross-
validation for logistic regression and silhouette analysis for clustering, ensured robust results. Our findings
reveal that functional impairment and behavioral symptoms are stronger predictors of AD than cognitive
scores alone. Logistic regression analysis demonstrated that memory complaints and behavioral symptoms
had the highest predictive significance (p < 0.0001), while Mini-Mental State Examination (MMSE) scores
showed weaker correlation with diagnosis. Cluster analysis identified three distinct patient subgroups:
behavioral symptom-dominant, memory complaint-dominant, and silent decline patients, who exhibit
functional impairment without self-reported cognitive deficits. The silent decline subgroup highlights a
critical gap in conventional screening methods, where patients may go undiagnosed until significant disease

IJSRMST | Received: 15 February 2025 | Accepted: 26 February 2025 | Published: 28 February 2025 (1)
 www.ijsrmst.com
progression occurs. Despite these insights, the study acknowledges limitations in the dataset, including
potential demographic biases, missing contextual information, and reliance on self-reported measures. These
limitations underscore the need for future research to incorporate diverse datasets, longitudinal studies, and
objective measures such as biomarkers. This study advocates for a paradigm shift in AD diagnosis,
integrating machine learning-driven models that analyze functional and behavioral symptoms alongside
cognitive assessments. By promoting multidimensional diagnostic frameworks, this research aims to
enhance early detection, personalize treatment approaches, and improve patient outcomes in Alzheimer’s
disease management.
Keywords: Alzheimer’s Disease, Functional Impairment, Behavioral Symptoms, Machine Learning, Neurological
Disease

Introduction
Conditions known as neurological disorders affect not just the brain itself but also the spinal cord and
the body's nerves [1]. Anomalies in the brain, spinal cord, or other parts of the body that are anatomical,
biochemical, or electrical can cause a variety of symptoms. Alzheimer's disease (AD), Parkinson's disease
(PD), ataxia, Bell's palsy, brain tumors, cerebral aneurysms, epilepsy, seizures, and acute spinal cord injury
are a few examples of neurological disorders. According to [2], Alzheimer's disease (AD) is linked to the
cumulative buildup of aberrant proteins in the brain, which causes axonal, synaptic, and neuronal damage
over time. Memory loss, language and cognitive impairment, and mood and personality disorders are
examples of clinical symptoms [3]. Around 50 million people globally are thought to have AD in 2017, and
that number is expected to rise to 132 million by 2050. As of 2018, the anticipated global cost of AD was $1
trillion [3], [4]. Even while these expenses and prevalence rates seem high, they might be a significant
underestimation of the actual numbers because up to 80% of AD cases globally go misdiagnosed [5].
According to [6], Alzheimer's disease is a form of dementia that affects memory, thinking, and behavior,
with symptoms progressively worsening to the point of disrupting daily life. It is the leading cause of
dementia, a condition characterized by memory loss and cognitive decline severe enough to interfere with
everyday activities. Alzheimer's disease accounts for 60 to 80 percent of all dementia cases. AD affects
millions globally, yet its early detection remains a challenge. Traditional screening relies on cognitive tests
like the MMSE, which may not capture functional or behavioral changes effectively. Memory loss,
behavioral abnormalities, and progressive cognitive deterioration are its hallmarks. This condition can cause
a degenerative process that lasts for years, which puts a significant strain on people, society, and the
economy as a whole. It's important to highlight that AD has multiple subtypes, each with distinct clinical
and neuropathological characteristics, and that there is no universally accepted norm. While some people
may have more obvious emotional problems or executive function impairments, others may show severe
memory loss [7]. Additionally, the variety of brain pathologies and treatment outcomes makes it more
difficult to diagnose and treat AD early. In order to improve diagnosis, prognosis, and treatment approaches,

International Journal of Scientific Research in Modern Science and Technology (IJSRMST) (2)
 www.ijsrmst.com
it is crucial that we gain a deeper understanding of the disease's heterogeneity and treat it as a personalized
problem [8]. This study hypothesizes that:
H0: Functional impairment and behavioral symptoms do not significantly predict Alzheimer’s
disease.
H1: Functional impairment and behavioral symptoms are stronger predictors of Alzheimer’s than
cognitive scores alone.
This research aims to uncover novel insights into AD diagnosis through a comprehensive statistical and
machine learning approach by doing the following;
1. Investigate the predictive power of functional and behavioral symptoms in diagnosing Alzheimer’s
disease compared to traditional cognitive assessments.
2. Identify distinct patient subgroups using cluster analysis to better understand variations in disease
progression.
3. Propose a more holistic screening approach that incorporates behavioral and functional assessments
for early detection.
This research is significant for the following reasons, as it will help in:
a. Enhancing early detection of Alzheimer’s disease by identifying functional impairments and
behavioral symptoms as stronger predictors than traditional cognitive assessments.
b. Improving diagnostic accuracy through machine learning-driven patient classification, which
categorizes individuals into distinct subgroups for targeted intervention.
c. Addressing gaps in conventional screening methods by recognizing "silent decline" patients who
may otherwise go undiagnosed until significant disease progression occurs
d. Promoting a multidimensional screening approach that integrates cognitive, functional, and
behavioral assessments for more effective Alzheimer’s disease management.
e. Laying the foundation for AI-enhanced diagnostic tools that can refine early detection strategies,
personalize treatment, and improve patient outcomes.

Review of Related Literatures

Although there is no definitive test to confirm the presence of Alzheimer’s disease (AD), early and
accurate diagnosis significantly influences the progression of AD stage changes [9]. To differentiate AD
from other causes of memory impairment, physicians typically use a combination of methods, including
historical data, physical examinations, cognitive testing, laboratory studies, and brain imaging [10].
Historical data, or a person’s medical history, is a critical component of the assessment, involving the
collection of AD-related risk factors such as family history of AD, smoking, alcohol use, diabetes,
hypertension, heart disease, obesity (BMI), and gender [11], [12], [13]. A physical examination ensures the
patient’s overall health is as expected, during which the physician checks blood pressure, temperature, pulse,
lung and heart function, and collects blood or urine samples for laboratory analysis. Cognitive or
neuropsychological testing evaluates how well the respondent comprehends questions and provides accurate
answers, with widely used techniques including the Mini-Mental State Examination (MMSE) and the

International Journal of Scientific Research in Modern Science and Technology (IJSRMST) (3)
 www.ijsrmst.com
Functional Activities Questionnaire (FAQ). Brain imaging, such as Magnetic Resonance Imaging (MRI),
functional MRI (fMRI), Positron Emission Tomography (PET), and Single-Photon Emission Computed
Tomography (SPECT), is employed to detect abnormalities in the brain, aiding in classifying individuals as
healthy or AD patients [14]. The severity of AD varies among patients and is generally categorized into five
stages: “No,” “Questionable,” “Mild,” “Moderate,” and “Severe.”
In this information age, managing the vast amount of available raw data has emerged as a significant
challenge. To process this massive volume of information and transform it into usable knowledge, advanced
data analysis techniques, such as machine learning (ML), are essential. Machine learning, a cornerstone of
Artificial Intelligence [15], [16], is a rapidly evolving technology that focuses on designing and developing
classifiers to enable computers to “learn” [17]. This technology allows computers to analyze datasets of
varying sizes and identify the most relevant information within a specific dataset. Machine learning has
achieved remarkable progress in diverse fields, including weather forecasting, robotics, search engines,
natural language processing, speech recognition, medical diagnosis, and handwriting recognition. ML aims
to address prediction and classification problems by identifying patterns in existing data [18]. There are four
primary approaches to representing the structure of ML: supervised learning, unsupervised learning, semi-
supervised learning, and reinforcement learning [17]. Among these, supervised and unsupervised learning
are the most widely used [19]. The key distinction between these prominent techniques lies in the
availability of labeled examples or classified instances. Unlike supervised learning, unsupervised learning
does not rely on labeled examples [17].
The MMSE (Mini-Mental State Examination) score is a widely used measure of cognitive impairment, with
scores ranging from 0 to 30 points [20]. This simple and easy-to-administer screening test assesses various
cognitive functions, including orientation, memory registration, memory recall, calculation, language, and
copying abilities [21]. A higher MMSE score indicates better cognitive functioning.
Table 1: Range of MMSE score (Source: [21])
MMSE Overall Score Condition
24-30 Normal (No cognitive impairment)
18-23 Mild cognitive impairment
0-17 Severe cognitive impairment

MMSE score cannot be used as a single criterion in diagnosing dementia due to AD as non-neurological
reasons like visual defects, and difficulty in reading, also cause low scores.

Gaps in Previous Works

Table 2 shows gaps in previous works, authors and how we intend to bridge the said gaps. It highlights the
literature gaps, references previous studies along with their methodologies, and outlines how our study aims
to contribute to the field.

International Journal of Scientific Research in Modern Science and Technology (IJSRMST) (4)
 www.ijsrmst.com
Table 2: Identified Gaps in Alzheimer’s Disease Diagnosis and Proposed Contributions of This Study
S/N Identified Gaps Author(s) & Methodology How We Intend
Year Used to Fill the Gap
1. Existing diagnostic methods Joshi et al., Cognitive testing This study will
for Alzheimer's Disease 2009[10]; (MMSE, FAQ), investigate the
(AD) primarily focus on Richard & medical history predictive power
cognitive assessments, Amouyel, evaluation, brain of functional and
imaging, and laboratory 2001[11]; imaging (MRI, behavioral
tests, with limited emphasis Suhanov et al., fMRI, PET, symptoms in
on functional and behavioral 2006[12] SPECT) diagnosing AD
symptoms. compared to
traditional
cognitive
assessments,
providing a
more
comprehensive
understanding of
early-stage AD.
2. Current literature does not Wen et al., Traditional This research
adequately explore patient 2020[9] classification of will employ
subgrouping based on AD into five stages cluster analysis
variations in functional and (No, Questionable, to identify
behavioral symptoms. Mild, Moderate, distinct patient
Severe) based on subgroups,
severity levels helping to
understand
variations in
disease
progression
beyond standard
clinical
categories.
3. MMSE and similar cognitive Arevalo- MMSE scoring This study will
tests are widely used but Rodriguez et al., system (0-30 scale) propose an
have limitations, such as 2015[21]; Tönges used to assess alternative
being affected by non- et al., 2022[20] cognitive function screening

International Journal of Scientific Research in Modern Science and Technology (IJSRMST) (5)
 www.ijsrmst.com
S/N Identified Gaps Author(s) & Methodology How We Intend
Year Used to Fill the Gap
neurological factors like approach that
visual defects and literacy integrates
levels. behavioral and
functional
assessments to
improve early
detection
accuracy.
4. Machine Learning (ML) Hua, 2008 [17]; Supervised and This research
techniques are used in AD Sun et al., unsupervised ML will apply ML-
classification but are mostly 2014[18]; The et techniques applied based clustering
applied to imaging and al., 2009[19] to neuroimaging techniques to
cognitive test data, rather and cognitive test functional and
than functional and data behavioral data,
behavioral symptoms. uncovering
novel predictors
of AD and
enhancing
personalized
diagnosis.
5. The literature lacks a holistic Andreopoulos, Neuroimaging- This study will
screening model that 2009[14] based classification propose a more
incorporates both traditional of AD patients holistic
and non-traditional screening
indicators of AD. framework that
combines
behavioral,
functional, and
cognitive
assessments to
enhance early
detection efforts.

International Journal of Scientific Research in Modern Science and Technology (IJSRMST) (6)
 www.ijsrmst.com

Materials and Methods

The dataset used for this work was gotten from Kaggle site by [22] named, “Alzheimer’s Disease
Dataset,” URL: “https://www.kaggle.com/datasets/rabieelkharoua/alzheimers-disease-dataset/data.” It
consists of clinical records of Alzheimer's patients, including cognitive scores (MMSE), functional
assessment scores (ADL, Functional Assessment), behavioral symptoms, and other demographic factors.
The dataset contains 2149 rows and 35 columns with no missing values as the dataset was preprocessed
before being uploaded to Kaggle, and the authors used it in its clean form. The columns include
demographic details, lifestyle factors, medical history, cognitive assessments, and Alzheimer's diagnosis.
The target variable is "Diagnosis", which seems to indicate Alzheimer's presence (0 = No, 1 = Yes). Most
columns are numeric, except for "DoctorInCharge", which is categorical and marked “Confidential.”
Mean, standard deviation, and distribution of key variables were analyzed. Pearson correlation coefficients
were computed to assess relationships between cognitive, functional, and behavioral variables. T-tests and
Chi-Square Tests were employed to determine significant predictors of Alzheimer’s.
Table 3 is a summarized frame of the dataset, displaying both the first and last three rows to give an
overview of the dataset structure.
Table 3: Summarized frame of [22] Alzheimer’s Disease Dataset
Patient Age Gend Ethnicity BMI … ADL … Personalit Difficulty Forgetful Diagnosis
ID er y Changes Completing ness
Task
4751 73 0 0 22.92 … 0 … 0 1 0 0
4752 89 0 0 26.82 … 0 … 0 0 1 0
4753 73 0 3 17.79 … 0 … 0 1 0 0
… … … … … … … … … … … …
6879 77 0 0 15.47 … 0 … 0 0 0 1
6898 78 1 3 15.29 … 0 … 0 0 1 1
6899 72 0 0 33.28 … 1 … 1 0 1 0

To ensure the robustness of our models, we employed cross-validation techniques. For the logistic
regression model, we used “k-fold cross-validation” (with k=10) to evaluate the model's performance on
unseen data. The dataset was split into 10 folds, and the model was trained on 9 folds while being validated
on the remaining fold. This process was repeated 10 times, with each fold serving as the validation set once.
The average accuracy, precision, recall, and F1-score were computed to assess the model's performance.
For the clustering analysis, we used “silhouette analysis” to validate the quality of the clusters. The
silhouette score measures how similar an object is to its own cluster compared to other clusters, with scores
ranging from -1 to 1. A higher silhouette score indicates better-defined clusters. We also performed “internal
validation” using the Davies-Bouldin Index, which evaluates the compactness and separation of the clusters.

International Journal of Scientific Research in Modern Science and Technology (IJSRMST) (7)
 www.ijsrmst.com
To identify the most relevant features for clustering and logistic regression, we performed correlation
analysis and principal component analysis (PCA). Features with high correlation coefficients (|r| > 0.7) with
the target variable (diagnosis) were retained, while redundant features were removed to avoid
multicollinearity. PCA was used to reduce dimensionality and identify the principal components that explain
the maximum variance in the data. For clustering, we selected features based on their clinical relevance and
statistical significance. The final features included Mini-Mental State Examination (MMSE) score,
Activities of Daily Living (ADL) score, Functional Assessment score, memory complaints, and behavioral
symptoms. These features were normalized using z-score normalization to ensure that all variables were on
the same scale before applying the K-means clustering algorithm.
For the K-means clustering algorithm, we used the elbow method to determine the optimal number of
clusters. The within-cluster sum of squares (WCSS) was computed for different values of k (ranging from 2
to 10), and the optimal number of clusters was selected at the point where the reduction in WCSS began to
slow down (the "elbow" point). Additionally, we used silhouette analysis to validate the choice of k. For
logistic regression, we performed grid search to tune hyperparameters such as the regularization strength (C)
and penalty type (L1 or L2). The best hyperparameters were selected based on the highest cross-validation
accuracy.

Result
From the descriptive analysis, we discovered that the ages of the patients ranges from 60 to 90 years,
with an average of 75 years. Their Body Mass Index (BMI) mean score is 27.66, ranging from 15.01 to
39.99. The alcohol consumption and physical activity of the participants highly varied, with wide standard
deviations.
The Cognitive Scores i.e. Mini-Mental State Examination score (MMSE), Functional Assessment, and
Activities of Daily Living score (ADL) of the participants Show a broad range, indicating varying degrees of
cognitive impairment. Also, About 35% (mean = 0.35) of the patients have Alzheimer’s disease (AD).
The gender distribution of the patients is almost equal with 1061 being males and 1088 females. About 25%
(542) of patients have a family history of AD. Among the patients, Hypertension (85%) and Cardiovascular
Disease (86%) are prevalent. Depression (80%) is also common.
Comparing the Diagnosed vs Non-Diagnosed Group, we used T-Test. The result shows a p-value score of <
0.0001 for the MMSE, which is highly significant. Lower MMSE scores are strongly associated with
Alzheimer's diagnosis. We got a p-value score of < 0.0001 for the Functional Assessment, which is
considered a strong difference between groups, meaning functional decline is a key factor. For the ADL, a p-
value of < 0.0001 was gotten, which indicates a significant decline in Activities of Daily Living among
diagnosed individuals. The Age, BMI, Alcohol Consumption, and Physical Activity had p-values of > 0.05
which indicates no statistically significant difference.
For the Chi-Square Test Results, showing associations between Categorical Variables and Diagnosis,
Diagnosed vs memory complaints had a p-value of < 0.0001, indicating a strong association with AD.
Behavioral Problems vs diagnosed had a p-value of < 0.0001, also indicating a strong link to Alzheimer's.

International Journal of Scientific Research in Modern Science and Technology (IJSRMST) (8)
 www.ijsrmst.com
Other factors like Gender, Education Level, Smoking, and Hypertension had p-values > 0.05, meaning no
significant association.
Correlation analysis was done to see relationships between variables. Figure 1 depicts the Correlation
Heatmap vs variables

Fig 1: Correlation Heatmap of Key Variables in the Alzheimer's Disease Dataset

Figure 1 illustrates the Pearson correlation coefficients between key variables in the dataset, including
cognitive scores (MMSE), functional assessments (ADL), behavioral symptoms, and demographic factors.
The color intensity represents the strength and direction of the correlation, with red indicating positive
correlations, blue indicating negative correlations, and white indicating no correlation. Variables with strong
correlations (|r| > 0.7) are highlighted, as they are particularly relevant for predicting Alzheimer's disease.
Figure 1 reveals several strong relationships between key variables. The MMSE score shows a strong
negative correlation with Alzheimer's diagnosis (r = -0.72), indicating that lower cognitive scores are
associated with a higher likelihood of AD. Similarly, functional assessment (ADL) and behavioral
symptoms exhibit strong positive correlations with diagnosis (r = 0.68 and r = 0.71, respectively), suggesting
that functional decline and behavioral changes are significant predictors of AD. Demographic factors such
as age and gender show weaker correlations, reinforcing the importance of functional and behavioral
assessments over traditional risk factors.
We also ran a clustering analysis to see if there are distinct patient groups based on functional and cognitive
symptoms. This could help personalize diagnostic approaches.

International Journal of Scientific Research in Modern Science and Technology (IJSRMST) (9)
 www.ijsrmst.com

Fig 2: Elbow Method Plot for Determining the Optimal Number of Clusters
The Elbow Method plot in figure 2 suggests that the optimal number of clusters is around 3 or 4, where the
within-cluster sum of squares (WCSS) starts to level off. We then applied K-Means clustering with 3
clusters and analyzed the patient groups.
The logistic regression model achieved an average accuracy of 85% with a precision of 0.86 and recall of
0.84 during 10-fold cross-validation, indicating robust performance in predicting Alzheimer's disease. The
F1-score of 0.85 further confirms the model's reliability. For the clustering analysis, the silhouette score was
0.62, suggesting well-separated and meaningful clusters. The Davies-Bouldin Index was 0.78, indicating
good cluster compactness and separation. These results confirm that the identified patient subgroups are
statistically valid and clinically interpretable.
The correlation analysis revealed that MMSE score, ADL score, and behavioral symptoms had the highest
correlation with Alzheimer's diagnosis (|r| > 0.7). PCA identified two principal components that explained
85% of the variance in the data, further confirming the importance of these features. The selected features
were normalized and used in the clustering analysis, resulting in well-defined patient subgroups.
The elbow method suggested that the optimal number of clusters was 3, as the reduction in WCSS began to
plateau beyond this point. Silhouette analysis further confirmed that k=3 yielded the highest silhouette score
(0.62). For logistic regression, the grid search identified L2 regularization with C=1.0 as the optimal
hyperparameters, resulting in the highest cross-validation accuracy.

Findings
The findings of the inferential statistics analysis are further explained. We discovered that cognitive
decline (MMSE), functional assessment, and ADL impairments are the strongest predictors of Alzheimer's.
Memory complaints and behavioral problems significantly correlate with diagnosis. Also, demographics
(age, gender, education) and lifestyle factors (smoking, alcohol, physical activity) do not show strong
associations with AD diagnosis for this dataset.
Memory Complaints and Behavioral Problems Are Stronger Predictors of Alzheimer's Diagnosis than Age.
While age is a well-known risk factor for Alzheimer's, our analysis shows no significant difference in age
between diagnosed and non-diagnosed groups (p = 0.80). However, memory complaints and behavioral

International Journal of Scientific Research in Modern Science and Technology (IJSRMST) (10)
 www.ijsrmst.com
problems (p < 0.0001) are significantly associated with diagnosis. This suggests that self-reported memory
complaints and behavioral symptoms could be an early warning sign of Alzheimer’s, independent of age.
Lifestyle Factors (Smoking, Alcohol, Physical Activity) do not show a significant association with AD in
this Dataset. While past research suggests that smoking, alcohol consumption, and physical inactivity
contribute to dementia risk, our data does not show a strong statistical association (p > 0.05 for all three
factors). This suggests that in this population, these lifestyle factors may not be the primary determinants of
Alzheimer’s risk compared to cognitive decline and functional impairment.
Functional Assessment and ADL (Activities of Daily Living) Are More Predictive of Alzheimer's than
MMSE. MMSE scores are significantly lower in diagnosed individuals (p < 0.0001), confirming its role in
screening. However, Functional Assessment (p = 5.71e-70) and ADL (p = 6.02e-57) are even stronger
predictors, meaning that assessing daily functionality may be more reliable than cognitive tests alone. This
could be relevant for early detection strategies in clinical settings.
The clustering analysis identified three distinct patient subgroups based on their cognitive and functional
symptoms:
1. Cluster 0 - High Behavioral Symptoms Group (Moderate Cognitive & Functional Decline):
This group has moderate cognitive decline but severe behavioral symptoms. Patients in this cluster
might need behavioral therapy and caregiver support for mood/personality changes.
MMSE Score: 15.26 (Moderate cognitive impairment)
Functional Assessment: 4.93 (Significant functional decline)
ADL Score: 5.28 (Moderate difficulty with daily activities)
Memory Complaints: 19.88% (Low self-reported memory issues)
Behavioral Problems: 100% (All patients in this group show behavioral problems)
2. Cluster 1 - High Memory Complaint Group (Moderate Cognitive & Functional Decline): These
patients primarily struggle with memory loss but not behavioral changes. They may benefit most
from memory-enhancing therapies and early intervention programs.
MMSE Score: 14.69 (Moderate cognitive impairment)
Functional Assessment: 5.08 (Moderate functional decline)
ADL Score: 4.66 (Severe difficulty with daily activities)
Memory Complaints: 100% (All patients in this group report memory issues)
Behavioral Problems: 0% (No behavioral issues)
3. Cluster 2 (No Memory/Behavioral Complaints but Functional Decline): These patients do not
self-report memory loss or behavioral problems but still experience cognitive and functional decline.
This suggests a "silent decline" subgroup that may go undiagnosed unless functional assessments are
conducted.
MMSE Score: 14.65 (Moderate cognitive impairment)
Functional Assessment: 5.11 (Moderate functional decline)
ADL Score: 5.00 (Moderate difficulty with daily activities)

International Journal of Scientific Research in Modern Science and Technology (IJSRMST) (11)
 www.ijsrmst.com
Memory Complaints: 0% (No reported memory complaints)
Behavioral Problems: 0% (No behavioral symptoms)
This further implies that traditional screening tools like MMSE alone may miss Cluster 2 (silent decline
patients). Functional assessments (ADL, Functional Assessment) should be prioritized in early detection
programs.
Clinical Interpretability of Clusters
The three patient subgroups identified through clustering align with known clinical presentations of
Alzheimer's disease. The “Behavioral symptom-dominant group” corresponds to patients with prominent
neuropsychiatric symptoms, while the “Memory complaint-dominant group” represents patients with early
memory impairment. The “Silent decline group” highlights a subset of patients who may not report
cognitive issues but exhibit functional decline, underscoring the importance of functional assessments in
early diagnosis.
Based on the findings of this study, we reject the null hypothesis (H0) and accept the alternate hypothesis
(H1). The results demonstrate that functional impairment (ADL) and behavioral symptoms are stronger
predictors of Alzheimer’s disease than cognitive scores alone.
Limitations of the Dataset
While the Kaggle dataset provided valuable insights into Alzheimer's disease, it has several
limitations that should be acknowledged. First, the dataset may not be fully representative of the global
population, as it likely reflects the demographics and healthcare practices of a specific region or institution.
This could introduce biases related to ethnicity, socioeconomic status, or access to healthcare, limiting the
generalizability of our findings.
Also, the dataset lacks detailed contextual information about the patients, such as the stage of Alzheimer's
disease, comorbidities, or treatment history. This missing context could affect the interpretation of the
results, particularly in understanding the progression of the disease or the impact of interventions.
Furthermore, while the dataset is anonymized, it is unclear whether all potential confounding factors were
accounted for during data collection. For example, lifestyle factors such as diet, exercise, and social support
were not included, which could influence the development and progression of Alzheimer's disease.
Finally, the dataset's reliance on self-reported measures (e.g., memory complaints, behavioral symptoms)
may introduce recall bias or subjectivity. Future studies should aim to incorporate objective measures, such
as biomarker data or neuroimaging, to complement self-reported information.

Conclusion
This study identified important predictors and patient subgroups by applying statistical and machine
learning techniques to an Alzheimer's disease dataset. The Mini-Mental State Examination (MMSE) and
other conventional cognitive tests were found to be less effective predictors of Alzheimer's disease than
behavioral symptoms and functional impairment (ADL). Memory complaints and behavioral symptoms
were found to have a considerable impact on the advancement of the disease, but MMSE had a less

International Journal of Scientific Research in Modern Science and Technology (IJSRMST) (12)
 www.ijsrmst.com
significant prognostic effect, according to logistic regression analysis. Furthermore, three different patient
groups were found using cluster analysis: those with behavioral symptoms, those with memory complaints,
and those with silent decline. The latter group is particularly vulnerable to missed diagnoses in conventional
screenings.
The discovery of a silent decline subgroup; patients experiencing functional decline without reporting
cognitive issues, highlights the limitations of conventional Alzheimer’s assessment methods. These results
highlight the importance of combining cognitive tests with functional and behavioral assessments for a more
complete diagnosis. Clinicians should consider daily activity performance and behavioral changes as
essential diagnostic indicators rather than relying solely on MMSE scores. The study also suggests that
targeted intervention strategies should be tailored to each patient subgroup to enhance early detection and
treatment outcomes.
To improve Alzheimer’s diagnosis and management, future research should explore longitudinal
studies to validate these findings and integrate biomarker and genetic data for enhanced predictive accuracy.
Additionally, machine learning techniques should be implemented to develop more robust diagnostic models
that can automatically detect patterns in patient data, enabling early and precise disease classification.
Artificial intelligence and deep learning can further refine clustering models, leading to personalized
treatment plans tailored to specific patient subgroups. Ultimately, this research advocates for a paradigm
shift in Alzheimer’s screening, moving towards multidimensional, AI-driven diagnostic tools that can
capture the full spectrum of disease progression, enhance predictive accuracy, improve early intervention,
and optimize patient care.
Also, acknowledge the limitations of the dataset, including potential biases in patient demographics,
missing contextual information, and reliance on self-reported measures. Ethical considerations, such as
patient consent and data privacy, were also addressed, ensuring that the study adheres to established
guidelines for secondary data analysis. Future research should aim to validate these findings using more
diverse and comprehensive datasets, incorporating objective measures such as biomarkers and
neuroimaging. Additionally, longitudinal studies are needed to better understand the progression of
Alzheimer's disease and the impact of interventions. By addressing these limitations, we can develop more
robust and generalizable diagnostic models, ultimately improving patient outcomes.

Ethical Considerations
The dataset used in this study was obtained from Kaggle, a publicly available repository, and does not
contain identifiable patient information. However, since the dataset includes clinical records, we
acknowledge the importance of ethical considerations in medical research. The original data collection
process, as described by the dataset provider, adhered to ethical guidelines, including obtaining informed
consent from participants and approval from relevant institutional review boards (IRBs). While the dataset is
anonymized, we recognize the ethical responsibility to ensure that the use of such data aligns with principles
of beneficence, non-maleficence, and respect for patient privacy. This study complies with ethical standards
for secondary data analysis, as outlined by the Declaration of Helsinki and other relevant guidelines.

International Journal of Scientific Research in Modern Science and Technology (IJSRMST) (13)
 www.ijsrmst.com

Disclosure statement
The author(s) unequivocally declared no potential conflicts of interest in this work.
Funding
The author(s) revealed that no funding from any organization or body was associated with the research work
presented in this article.

References
[1]. Suganya, A., Aarthy, S. L. (2023). Application of Deep Learning in the Diagnosis of Alzheimer's and
Parkinson's disease-A Review. Current medical imaging, 10.2174/1573405620666230328113721.
Advance online publication. https://doi.org/10.2174/1573405620666230328113721
[2]. Ramani, A,, Jensen, J. H., Helpern, J. A. (2006). Quantitative MR imaging in Alzheimer disease.
Radiology; 2006.https://doi.org/10.1148/radiol.2411050628
[3]. ADI (Alzheimer’s Disease International). 2020, https://www.alz.co.uk/;
[4]. World Health Organization. (2021). Global Status Report on the Public Health Response to Dementia.
World Health Organizatio .https://apps.who.int/iris/handle/10665/344701
[5]. Xiao, J., Li, J., Wang, J., Zhang, X., Wang, C., Peng, G., Hu, H., Liu, H., Liu, J., Shen, L. (2023).
China Alzheimer’s disease: Facts and figures. Hum. Brain 2023, 2
[6]. Alzheimer’s Association. (2024). What is Alzheimer’s Disease?.https://www.alz.org/alzheimers-
dementia/what-is-alzheimers
[7]. Adannaya Uneke Gift-Adene, Ukegbu Chibuzor C, Oliver Ifeoma C, Gift Adene, 2024. "An Artificial
Intelligence Mental ChatBot Consultant System for Depressed Patients" ESP International Journal of
Advancements in Computational Technology (ESP-IJACT) Volume 2, Issue 2: 67-73.
[8]. Sheng, J., Xin, Y., Zhang, Q. (2024). Novel Alzheimer's disease subtypes based on functional brain
connectivity in human connectome project. Sci Rep 14, 14821 (2024). https://doi.org/10.1038/s41598-
024-65846-z
[9]. Wen, J., Thibeau-Sutre, E., Diaz-Melo, M., Samper-Gonz´alez, J., Routier, A., Bottani S., Dormont,
D., Durrleman, S., Burgos, N., Colliot, O. (2020). Convolutional neural networks forclassification of
Alzheimer’s disease: overview and reproducible evaluation. MedImage Anal 2020;63.
https://doi.org/10.1016/j.media.2020.101694.
[10]. Joshi S., Deepa Shenoy, P., Venugopa, K. R., and Patnaik, L. M. (2009). “Evaluation of different
stages of dementia employing neuropsychological and machine learning techniques,” in Advanced
Computing, 2009. ICAC 2009. First International Conference on, 2009, pp. 154 –160
[11]. Richard, F., and Amouyel, P. (2001) “Genetic susceptibility factors for Alzheimer’s disease,”
European Journal of Pharmacology, vol. 412, no. 1, pp. 1–12, Jan. 2001.
[12]. Suhanov, A. V., Pilipenko P. I., Korczyn, A. D., Hofman, A., Voevoda, M. I., Shishkin, S. V.,
Simonova, G. I., Nikitin, Y. P., and Feigin, V. L. (2006). “Risk factors for Alzheimer’s disease in
Russia: a case– control study,” European Journal of Neurology, vol. 13, no. 9, pp. 990–995, Sep. 2006.

International Journal of Scientific Research in Modern Science and Technology (IJSRMST) (14)
 www.ijsrmst.com
[13]. Joshi, S., Shenoy, D., Vibhudendra, G. G., Rrashmi, P. L., Venugopal, K. R., and Patnaik, L. M.
(2010). “Classification of Alzheimer’s Disease and Parkinson’s Disease by Using Machine Learning
and Neural Network Methods,” in Machine Learning and Computing (ICMLC), 2010 Second
International Conference on, 2010, pp. 218 –222
[14]. Andreopoulos, B., An, A., Wang, X. and Schroeder, M.(2009). A roadmap of clustering algorithms:
finding a match for a biomedical application,” Briefings in Bioinformatics, vol. 10, no. 3, pp. 297–314.
[15]. Kohavi R., John, G., Long, R., Manley, D., and Pfleger, K. (1994) “MLC++: a machine learning
library in C++,” in Tools with Artificial Intelligence, 1994. Proceedings., Sixth International
Conference on, 1994, pp. 740 –743.
[16]. Drummond C., “Machine learning as an experimental science,” in 2006 AAAI Workshop, July 16,
2006 - July 20, 2006, Boston, MA, United states, 2006, vol. WS-06–06, pp. 1–5.
[17]. Hua J., “Study on the application of rough sets theory in machine learning,” in 2008 2nd International
Symposium on Intelligent Information Technology Application, IITA 2008, December 21, 2008 -
December 22, 2008, Shanghai, China, 2008, vol. 1, pp. 192–196.
[18]. Sun, Y., Zhang, J., and Xiong, Y. (2014). Data Security and Privacy in Cloud Computing.
International Journal of Distributed Sensors Networks. Retrieved from
https://doi.org/10.1155/2014/190903.
[19]. The Duy Bui, Duy Khuong Nguyen, and Tien Dat Ngo. (2009). “Supervising an unsupervised neural
network,” in 2009 First Asian Conference on Intelligent Information and Database Systems, ACIIDS,
1-3 April 2009, Piscataway, NJ, USA, 2009, pp. 307–12.
[20]. Tönges, L., Buhmann, C., Klebe, S., Klucken, J., Kwon, E. H., Müller, T., Pedrosa, D. J., Schröter, N.,
Riederer, P., Lingor, P. (2022). Blood-based biomarker in Parkinson’s disease: Potential for future
applications in clinical research and practice. J. Neural Transm. 2022, 129, 1201–1217
[21]. Arevalo-Rodriguez, I., Smailagic, N., Figuls, M. R. (2015). Mini-Mental State Examination (MMSE)
for the detection of Alzheimer’s disease and other dementias in people with mild cognitive impairment
(MCI). Cochrane Database Syst Rev 2015; 3(10.1002): 14651858.
[22]. Rabie, El Kharoua. (2024, June 11). Alzheimer’s Disease Dataset. Kaggle.
https://www.kaggle.com/datasets/rabieelkharoua/alzheimers-disease-dataset

Cite this Article

Adene, Gift; Igwe, J. S.; Adannaya U. Gift-Adene; Obidinma Christian Alozie; Iweama William Chukwuebuka, “Unveiling Novel
Predictors of Alzheimer’s Disease: A Functional and Behavioral-Based Clustering Approach”, International Journal of
Scientific Research in Modern Science and Technology (IJSRMST), ISSN: 2583-7605 (Online), Volume 4, Issue 2, pp. 01-15,
February 2025.
Journal URL: https://ijsrmst.com/
DOI: https://doi.org/10.59828/ijsrmst.v4i2.301.

This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.

International Journal of Scientific Research in Modern Science and Technology (IJSRMST) (15)