Addressing Educational Disparities: Assessing the Gap for Indigenous

Community

Shafaq Khan, Viutika Rathod, Abhirup Ranjan, Anika Anjum Una and Neel Manish Pandya

School of Computer Science, University of Windsor, Ontario, Canada

Keywords:

Indigenous Knowledge, Indigenous Perspectives, Educational Disparities, Indigenous Students, Database.

Abstract:

This research paper explores the integration of Indigenous knowledge and perspectives in education to ad-

dress the educational disparities faced by Indigenous students in Canada [Change.org, 2023]. It proposes a

management system utilizing a database and logistic regression model to predict student dropout rates based

on key factors such as Cultural Identity, Gender, Government Funding, and more [Government of Canada,

2021]. The logistic regression model achieved an accuracy rate of 0.831 on the testing dataset and provided

valuable insights into the factors inﬂuencing dropout rates among Indigenous students. The paper emphasizes

the importance of cultural sensitivity, ethical considerations, and collaboration with Indigenous communities

throughout the research process. While logistic regression offers interpretability and simplicity, future work

may explore the use of other machine learning models and qualitative data to enhance accuracy and gain

deeper insights. The goal is to promote educational equity and inclusivity while respecting Indigenous knowl-

edge and aspirations in Canadian education. This will help the local and federal authorities in determining

and early forecasting that the student might drop out and accordingly the authorities can take action. Also the

data and results from different regions can be help in determining how they are performing and if their action

planning can be implemented in the other region as well to reduce the drop out ratio. This will help in the

overall development of the indigenous community.

1 INTRODUCTION

The Indigenous communities in Canada experience

signiﬁcantly higher dropout rates compared to non-

Indigenous communities in this part of the world, at-

tributed to a range of factors such as historical and

inter-generational impacts, ﬁnancial inequities, so-

cial marginalization, lack of support systems, and

geographic barriers (Government of Ontario, 2017).

These factors restrict access to quality education for

Indigenous students, resulting in lower educational

achievements compared to the general population of

Canada (CFS Ontario, 2021).

Outline of the problem of educational disparities

faced by Indigenous students in Canada, there are

multiple factors such as the legacy of colonialism,

residential schools, racism, and insufﬁcient funding

hinder access to post-secondary education, which is

recognized as a treaty right for Indigenous students.

Funding inadequacy and access difﬁculties contribute

to a signiﬁcant disparity in educational opportuni-

ties (Brown, 2023). Also, the lack of necessary in-

frastructure has hindered the distribution of cultur-

ally relevant educational curricula to Indigenous com-

munities. Teaching practices in non-Indigenous in-

stitutions need reﬁnement, focusing on incorporating

Indigenous history, cultures, and perspectives, and

addressing racism and marginalization (Government

of Canada, Interagency Advisory Panel on Research

Ethics, 2019).

Our work in this research paper is motivated by

a deep understanding of past injustices and current

hardships faced by Indigenous communities. We aim

to address the lack of cultural responsiveness in the

mainstream educational system and promote justice,

fairness, and reconciliation (Weston, 2019).Our goal

is to assist the government authorities of Canada in

understanding the dropout rates of Indigenous stu-

dents so that they can devise appropriate strategies for

their advancement and development. By embracing

empathy and recognizing the generational effects of

policies like the residential school system, we seek to

restore pride, dignity, and self-respect among Indige-

nous children (Kim, 2019). Through cultural respon-

siveness, we aim to foster respect, understanding, and

inclusive education for all students.

Our objective here is to bridge the educational gap

between Indigenous and non-Indigenous populations

in Canada by addressing issues such as insufﬁcient

funding, cultural disconnection, discrimination, and

Khan, S., Rathod, V., Ranjan, A., Una, A. and Pandya, N.

Addressing Educational Disparities: Assessing the Gap for Indigenous Community.

DOI: 10.5220/0012253600003693

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 16th International Conference on Computer Supported Education (CSEDU 2024) - Volume 2, pages 453-460

ISBN: 978-989-758-697-2; ISSN: 2184-5026

453

lack of support systems for Indigenous communities.

We aim to assess and analyze the quality of educa-

tion for Indigenous students, promote evidence-based

decision making, and ensure educational equity. In-

tegrating Indigenous knowledge systems, languages,

and histories into the curriculum is a crucial focus,

empowering Indigenous students and improving the

educational experience for all.

To achieve our goals, we propose a model that

utilizes a database to address educational disparities

faced by Indigenous students. Also using the Logis-

tic Model we will capture unique information such as

cultural identity, language proﬁciency, community in-

volvement, gender, and other relevant details, while

adhering to Indigenous data governance principles.

This will assist us in determining whether a particular

person may drop out, and if so, it will enable con-

cerned local or governmental bodies to take appro-

priate action. It will also enable comprehensive data

integration from various sources, evaluating the suc-

cess of educational initiatives and facilitating timely

support and prevention of widening educational gaps.

There is a lack of published papers addressing our

identiﬁed issue, making our research work unique.

Through our comprehensive model and data-driven

approach, we aim to contribute to the development

of effective strategies and policies that empower In-

digenous students and bridge the educational gap in

Canada.

2 BACKGROUND STUDY

The paper by Wu, J., Lin, S., Kong, H., and Shi, H.

(2019) titled ”The Combination Forecasting Model of

Telecommunication User Tricking Account Overdraft

Limit Based on Logistic Regression and SVM ex-

plores the role of educators, particularly those teach-

ing aspiring conservation practitioners, in responding

to the Truth and Reconciliation Commission (TRC)

and the National Inquiry into Missing and Murdered

Indigenous Women and Girls (MMIWG). The em-

phasis is on the signiﬁcance of reconciliation and eth-

ical engagement with Indigenous Peoples through a

revolutionary approach to teaching indigenous knowl-

edge. The focus is on fostering understanding, empa-

thy, and respect for Indigenous knowledge and aspi-

rations while equipping students with critical analysis

skills for ethical engagement. The paper calls for uni-

versities to ”Indigenize” their approaches, embracing

anti-racism, humility, reciprocity, and confronting on-

going colonialism and white supremacy. The goal is

to create a learning environment that respects and val-

ues Indigenous scholars, knowledge,and voices, fos-

tering re-conciliatory relationships between conser-

vation practitioners and Indigenous Peoples. Efforts

should go beyond course content and focus on build-

ing an anti-racist, anti-oppressive campus culture that

centers Indigenous perspectives and enables Indige-

nous intellectual expression. The paper advocates for

hiring more Indigenous faculty members and center-

ing Indigenous Peoples as experts about themselves

in the curriculum. The goal is to prepare students

to engage with Indigenous Peoples in a just and af-

ﬁrming manner while respecting Indigenous knowl-

edge and aspirations. The possible difﬁculties of

successfully integrating Indigenous knowledge within

the current curriculum and ensuring cultural sensitiv-

ity and authenticity in its application might, however,

be a drawback of this strategy (Wu et al., 2019).

In this study by Li, J., Brar, A., and Roihan, N.

(2021) “The Use of Digital Technology to Enhance

Language and Literacy Skills for Indigenous People:

A Systematic Literature Review,”the focus is on the

use of digital technologies to support Indigenous peo-

ple’s language and literacy learning, particularly in

English. Here, the emphasis is on addressing the neg-

ative inter-generational impacts of colonization and

socioeconomic stress on Indigenous academic perfor-

mance. The systematic review of 25 empirical stud-

ies provides insights into the efﬁcacy of digital tech-

nology in supporting Indigenous learners. While the

studies demonstrate positive outcomes, there are limi-

tations, such as the lack of rigorous research methods

and comprehensive reporting. To improve the effec-

tiveness of digital technology-based interventions, fu-

ture research should consider culturally relevant mul-

tiliteracies frameworks, engage in longitudinal stud-

ies to track students’ progress and incorporate more

Indigenous cultural elements. Additionally, there is a

need for data coding schemes to capture the nuances

of Indigenous language and literacy learning. The

research emphasizes the importance of culturally re-

sponsive practices, partnership with Indigenous com-

munities, and addressing the unique contextual fac-

tors affecting Indigenous education. The reliance on

digital infrastructure and access to technology, how-

ever, may be a weakness of their strategy or approach

and present difﬁculties in isolated or disadvantaged

Indigenous communities where dependable internet

connectivity may be scarce (Li et al., 2021).

Exploring the NOW (Northern Oral Language and

Writing) Play project, Stagg Peterson, S., and Dwyer,

B. (2016) in their paper titled ”Research in Canada’s

Northern Rural and Indigenous Communities: Sup-

porting Young Children’s Oral Language and Writ-

ing” shed light on efforts to enhance oral language

and writing skills among young children in north-

CSEDU 2024 - 16th International Conference on Computer Supported Education

454

ern rural and Indigenous communities. The research

emerged from the concerns raised by kindergarten

teachers in northern Canadian communities about stu-

dents’ limited language abilities upon entering school.

The project involved collaborating with public school

divisions in small rural communities and local educa-

tion authorities in Indigenous regions of four Cana-

dian provinces. Focusing on the assessment and sup-

port of young children’s oral language and writing

development in play contexts. These communities

face challenges related to geographic isolation, lim-

ited teaching resources, and a lack of opportunities for

teachers’ professional learning. The project empha-

sized the use of play contexts to enhance children’s

language and literacy skills. Additionally, the project

aimed to develop a culturally relevant oral language

assessment tool. This tool was designed to capture

the diverse ways in which children use language for

various social purposes during play and small-group

academic activities. Teachers and researchers collab-

oratively analyzed video recordings and transcripts

of children’s play to identify different language uses

and track individual children’s language development

(Stagg Peterson and Dwyer, 2016). However, the

project also acknowledges challenges such as parental

resistance to play based approaches and the necessity

for cultural sensitivity in research methods. Careful

navigation of these obstacles is crucial to ensure the

project’s ultimate success and positive impact.

The paper, ”Educational Equity in Canada: The

Case of Ontario’s Strategies and Actions to Ad-

vance Excellence and Equity for Students” (Camp-

bell, 2020) by Carol Campbell, delves into the edu-

cational equity initiatives in Ontario, Canada, focus-

ing on two main strands: the Literacy and Numer-

acy Strategy and the Student Success/Learning to 18

Strategy. With a commitment to multiculturalism and

diverse outcomes, the paper discusses policies tar-

geting gender, language learners, special education

needs, and Indigenous students within the context of

two speciﬁc policy strands. While the literature re-

view provides a comprehensive historical perspective,

it could beneﬁt from a more critical analysis of the

strategies’ effectiveness, including a nuanced explo-

ration of Indigenous issues and a discussion of poten-

tial limitations. The paper considers students across

K-12 levels but could be strengthened by specifying

the age groups within this range.

Moules et al.’s (2014)(Schaub, 2020) research

delves into the educational disparity between Indige-

nous and non-Indigenous students in Canada, em-

ploying hermeneutics for an in-depth understanding.

Referring to it as an ”educating gap,” the study em-

phasizes the potential for actionable solutions. Ex-

amining historical issues in Indigenous education, the

paper advocates for meaningful change while recog-

nizing leadership challenges. The investigation ex-

plores the gap from various perspectives, including

quantitative data, personal narratives, and perceptual

insights. Despite the strengths in its approach, the

study could strengthen its discussion by explicitly ad-

dressing limitations and providing a detailed analysis

of the term ”deﬁcit model” in education. The authors

acknowledge constraints such as time, ﬁnancial re-

sources, personal boundaries, and regional focus in

coastal British Columbia, expressing conﬁdence in

the broader applicability of their ﬁndings.

The studies by Wu et al. (2019), Li et al. (2021),

Stagg Peterson and Dwyer (2016),Campbell (2020)

and Moules et al’s(2014) collectively focus on the in-

tegration of Indigenous knowledge across various ed-

ucational contexts. While acknowledging challenges

and limitations in implementing this integration goal,

these papers emphasize the importance of reconcil-

iatory relationships, culturally responsive practices,

and addressing unique contextual factors affecting In-

digenous education. Notably, it should be highlighted

that the background studies, speciﬁcally concentrate

on younger students, offering insights into enhancing

language and literacy skills among children in north-

ern rural and Indigenous communities.

3 PROPOSED MODEL

In this section, we propose a predictive model us-

ing Logistic Regression, a binary classiﬁcation algo-

rithm. The model predicts whether a student will drop

out (1) or not (0) by using the key factors that have

been identiﬁed as input variables. Its interpretable re-

sults allow for a clear understanding of the impact of

predictor variables on dropout rates, helping to iden-

tify key factors contributing to the disparities (Das,

2021). Moreover, logistic regression provides prob-

ability estimates, prioritizing higher-risk students for

support systems. With its low complexity, minimal

data preprocessing requirements, and ease of imple-

mentation, logistic regression offers a practical and

efﬁcient solution for handling the data management

system. While it may not capture complex relation-

ships as effectively as advanced algorithms, combin-

ing it with other techniques through ensembling could

further enhance its predictive performance and lead to

more comprehensive strategies for bridging the edu-

cational gap among Indigenous communities.

The proposed model takes a systematic approach

to predicting the likelihood of dropout among Indige-

nous students (as shown in Figure 1). It starts by

Addressing Educational Disparities: Assessing the Gap for Indigenous Community

455

compiling pertinent data from various sources, such

as governmental agencies and academic institutions

(Bisong, 2019).

Figure 1: Workﬂow diagram.

Key traits like cultural identity, language proﬁ-

ciency, and educational level are chosen after the data

has been preprocessed. Following that, the dataset is

divided into training and test sets. To determine the

correlation between the selected features and dropout

outcomes, the logistic regression model is trained us-

ing the training data. The model is tested on the test-

ing set, and metrics like accuracy, precision, recall,

F1-score, and ROC-AUC (K, 2020) are used to as-

sess its performance. An effective predictive model

for comprehending dropout patterns among Indige-

nous students is being developed through this thor-

ough process.

4 METHODOLOGY AND

EXPERIMENTATION

The logistic regression model was implemented as

part of our proposed management system to address

educational disparities faced by Indigenous students

in Canada. The model’s objective was to predict

the likelihood of student dropout (1) or non-dropout

(0) based on key factors, including Cultural Identity,

Gender, Government Funding, Type of Educational

Institute, Employment Sector, Language Proﬁciency,

Community Involvement, Age, and Level of Educa-

tion. We trained the model using a binary classiﬁ-

cation algorithm and evaluated its performance using

various metrics.

4.1 Data Collection

A variety of sources, including the Alberta Open

Data Portal, Statistics Canada, the FNIGC (First Na-

tions Information Governance Center), and the Gov-

ernment of British Columbia, contributed to the col-

lection of data for preprocessing in Indigenous edu-

cation research. The dataset, consisting of indige-

nous headcount enrolment within the Alberta Post

Secondary Education-system with a sample size of

10000 records, includes various types of information,

such as demographic data (e.g., age, gender), cultural

identity, government funding status, enrollment ﬁg-

ures, and academic performance metrics. To ensure

accuracy and suitability for analysis, the data under-

went rigorous cleaning, transformation, and organi-

zation processes. Outliers were addressed, features

were normalized or encoded as required, and missing

values were imputed. Such pre-processing steps are

crucial, especially with a dataset of this size and com-

plexity, to prepare the data for subsequent analysis,

such as logistic regression modeling. This modeling

aims to shed light on the variables inﬂuencing the aca-

demic performance and dropout rates of Indigenous

students, providing valuable insights into the educa-

tional landscape.

The Figure 2 illustrates the credential they have

achieved so far in the Alberta region based on the

dataset available on Alberta Open Data Portal web-

site.

Figure 2: Distribution of Indigenous students based on their

education level in Alberta Region.

4.2 Data Processing

For data preprocessing, we will be categorizing the

ﬁelds into two categories: Binary and Numerical

variables (Rao et al., 2023). After one-hot encoding,

there will be only two values for categorical variables

which we will include as binary variables. Here is the

categorization:

Binary Variables

• Cultural Identity (after binary encoding)

– First Nations (1 for ”First Nations”, 0 for oth-

ers)

– Inuit (1 for ”Inuit”, 0 for others)

– M

etis (1 for ”M

etis”, 0 for others)

• Gender (after binary encoding)

CSEDU 2024 - 16th International Conference on Computer Supported Education

456

– Gender (1 for ”Female”, 0 for ”Male”)

• Government Funding (after binary encoding)

– Government Funding (1 for ”Yes”, 0 for ”No”)

Numerical Variables

• Language Proﬁciency (after label encoding)

– Language Proﬁciency (0 for ”Basic”, 1 for ”In-

termediate”, 2 for ”Advanced”)

• Community Involvement (after label encoding)

– Community Involvement (0 for ”Low”, 1 for

”Medium”, 2 for ”High”)

• Age (Numerical variable)

– Age (numeric values)

• Level of Education (after label encoding)

– Level of Education (0 for ”School”, 1 for ”High

School”, 2 for ”Bachelor’s”, 3 for ”Master’s”,

etc.)

• Type of Educational Institute (after label en-

coding)

– Type of Educational Institute (0 for ”Public

School”, 1 for ”Private School”, 2 for ”Home-

schooling”, 3 for ”Online Learning”)

• Employment Sector (after label encoding)

– Employment Sector (0 for ”Government”, 1 for

”Private”, 2 for ”Others”)

Table 1: Summary of Categorical and Numerical Features

in the Dataset.

Field Category

First Nations Binary (Cultural Identity)

Inuit Binary (Cultural Identity)

etis Binary (Cultural Identity)

Gender Binary

Government Funding Binary

Language Proﬁciency Numerical (Ordinal)

Community Involvement Numerical (Ordinal)

Age Numerical (Ordinal)

Level Of Education Numerical (Ordinal)

Type of Educational Insti-

tute

Numerical (Ordinal)

Employment Sector Numerical (Ordinal)

4.3 Feature Selection

Recursive feature elimination (RFE) (Misra and Ya-

dav, 2020) is used in the proposed algorithm for

addressing educational disparities among Indigenous

students in Canada to pinpoint the key characteristics

that signiﬁcantly inﬂuence the likelihood of dropout

prediction made by logistic regression. Initially, the

logistic regression model is trained using all perti-

nent features, including binary and numerical vari-

ables. Following a systematic elimination of less sig-

niﬁcant features, the RFE process ranks the remain-

ing features according to how well they predict aca-

demic outcomes. The subset of crucial elements with

the greatest inﬂuence on Indigenous students’ dropout

rates is identiﬁed through this iterative process.

Algorithm 1: Algorithm: Feature Selection using RFE and

Logistic Regression.

Data: Input data X, Target variable y

Result: Selected features X

selected

Encode categorical features;

enc

← OneHotEncode(X);

Split data into training and testing sets;

train

test

train

test

←

split(X

enc

,y,test size = 0.2, random state =

42,stratify = y);

Create logistic regression model with

balanced class weights;

model ← LogisticRegression(class weight =

’balanced’);

Implement RFE for feature selection;

num features ← 5;

r f e ← RFE(model,n features to select =

num features);

r f e.ﬁt(X

train

);

Fit model on selected features;

train sel

← X

train

[:,r f e.support ];

model.ﬁt(X

train sel

train

);

Predict target variable on the testing set;

test sel

← X

test

[:,r f e.support ];

pred

← model.predict(X

test sel

);

Calculate F1-score and accuracy;

f 1 ← F1-score(y

test

pred

);

accuracy ← accuracy score(y

test

pred

);

Print results;

print(”F1-score for ’Dropout’:”, f1);

print(”Accuracy:”, accuracy);

The F1-score is used in the above algorithm to

evaluate how well the logistic regression model pre-

dicted educational outcomes for Indigenous students.

Particularly when working with datasets that are un-

balanced, it aids in assessing the model’s capacity to

balance precision and recall. The computational com-

plexity of RFE, which is roughly O(np

/2), is also

taken into account when choosing the features. With

RFE’s iterative design, subsets of features are used to

train the logistic regression model, allowing for the

effective identiﬁcation of key predictors to reduce ed-

ucational disparities among Indigenous communities.

Addressing Educational Disparities: Assessing the Gap for Indigenous Community

457

4.4 Data Split

The ”train-test split,” in which the dataset is divided

into two parts: “a training set” and “a test set”, is the

most popular data splitting technique.

1. The training set and the test set should be sep-

arated from the dataset. The logistic regression

model will be trained using the training set, and

its performance will be assessed using the test set.

The training set is utilized to train the machine

learning model. During this phase, the model

learns patterns and relationships within the data.

The test set is reserved for evaluating the model’s

performance. It consists of data that the model has

not seen during the training phase, allowing for an

unbiased assessment of its generalization capabil-

ity.

2. The split ratio is typically 80% training data and

20% test data, but it can be changed depending

on the size of the dataset and the particular re-

quirements.Given our dataset size of 10000, we

have allocated 8000 records for training and 2000

records for testing.

4.5 Model Training

In order to accurately predict whether a student is

likely to drop out (1) or not (0) based on the selected

features: ”Cultural Identity,” ”Gender,” ”Government

Funding,” ”Type of Educational Institute,” ”Employ-

ment Sector,” ”Language Proﬁciency,” ”Community

Involvement,” ”Age,” and ”Level of Education,” the

logistic regression model must be trained to recog-

nize patterns and relationships within the training data

(Parker et al., 2013).

For each data point in the training set, the calcu-

lated values in the context of logistic regression re-

fer to the linear combinations of the chosen features

(’Cultural Identity’, ’Gender’, ’Government Fund-

ing’, ’Type of Educational Institute’, ’Employment

Sector’, ’Language Proﬁciency’, ’Community In-

volvement’, ’Age’, ’Level of Education’). For each

data point, the model will compute a weighted sum

of the feature values and add a bias term. The linear

combination (y) for a single data point (x) is com-

puted mathematically as follows:

y = b

+ b

+ . .. + b

(1)

Here:

• y is the calculated value for a given data point.

• b0 is the bias term (intercept).

• b1, b2, ..., bn are the weights (coefﬁcients) as-

signed to each feature.

• x1, x2, ..., xn are the feature values for the corre-

sponding data point.

The calculated values (y) will then be subjected to

the sigmoid function to convert them into probabili-

ties. The formula for the function is:

p =

1 + exp(−y)

(2)

Figure 3: The S Curve: Visualization of Logistic Regres-

sion’s Probabilistic Model.

Thus, p is the logistic regression model’s out-

put—the likelihood that a student will drop out—for

a speciﬁc data point.

The exponential representation of -y is represented by

exp

−y

. No matter the range of the calculated val-

ues (y), the sigmoid function (as shown in Figure 3

(Essampally, 2020)) will make sure that the output

probabilities (p) are bounded between 0 and 1. For

the corresponding data point, the model will predict a

dropout (1) when p is greater than or equal to 0.5, and

a non-dropout (0) when p is less than or equal to 0.5.

The regression algorithm will iteratively adjust the

weights (b1, b2,..., bn) and the bias term (b0) during

the model training process using optimization tech-

niques like gradient descent. The goal is to identify

the best combination of weights and biases to mini-

mize the discrepancy between the predicted probabil-

ities and the actual binary outcomes (dropout or non-

dropout) in the training data, thereby enhancing the

model’s capacity to make precise predictions on fresh,

untested data.

4.6 Model Evaluation

Evaluation metrics like accuracy, precision, recall,

F1-score, and ROC-AUC (Area Under the Curve - Re-

ceiver Operating Characteristic) are used to assess the

performance of the logistic regression model, which

was trained using the features like Cultural Identity,

Gender, Government Funding, Type of Educational

Institute, Employment Sector, Language Proﬁciency,

Community Involvement, Age, and Level of Educa-

tion. These metrics evaluate how well the model per-

forms in accurately predicting Indigenous students’

CSEDU 2024 - 16th International Conference on Computer Supported Education

458

likelihood of dropping out of school. We can assess

the model’s efﬁcacy in addressing educational dispar-

ities and promoting educational equity among Indige-

nous communities in Canada by analyzing these eval-

uation results on the testing dataset.

4.7 Interpretability

The logistic regression model’s feature importance

(coefﬁcients) offers important insights into how each

feature affects the likelihood that Indigenous students

will drop out of school. While negative coefﬁcients

suggest factors linked to lower dropout rates, pos-

itive coefﬁcients point to factors linked to higher

dropout rates. Stakeholders can pinpoint key causes

of educational disparities by looking at these coef-

ﬁcients for categories like Cultural Identity, Gender,

Government Funding, Type of Educational Institute,

Employment Sector, Language Proﬁciency, Commu-

nity Involvement, Age, and Level of Education. In

order to promote educational equity for Indigenous

communities, targeted interventions and support sys-

tems are informed by this understanding. While there

are other interpretability techniques, Feature Impor-

tance (Coefﬁcients) stands out for its clarity, usabil-

ity, and universal interpretability, enabling evidence-

based decision-making to effectively close the educa-

tional gap for Indigenous students.

5 RESULTS

The logistic regression model demonstrated promis-

ing results in predicting dropout rates among Indige-

nous students. The model achieved an accuracy rate

of 0.831 on the testing dataset, indicating its ability

to correctly classify dropout outcomes. Moreover, the

F1 score was 0.831 (as shown in Figure 4), showcas-

ing its effectiveness in identifying students at risk of

dropping out and minimizing false positives.

Figure 4: Showing the Accuracy for Dropped out instances

for Indigenous Students.

The interpretability of the logistic regression

model allowed us to identify key factors inﬂuencing

dropout rates among Indigenous students. The coefﬁ-

cients for various features provided valuable insights

into the impact of Cultural Identity, Gender, Gov-

ernment Funding, Type of Educational Institute, Em-

ployment Sector, Language Proﬁciency, Community

Involvement, Age, and Level of Education on edu-

cational outcomes. For instance, positive coefﬁcients

indicated factors associated with higher dropout rates,

while negative coefﬁcients pointed to factors linked

to lower dropout rates. This understanding empow-

ered us to design targeted interventions and support

systems to enhance educational equity and bridge the

educational gap for Indigenous students in Canada.

6 LIMITATIONS/CHALLENGES

The proposed approach to addressing educational dis-

parities faced by Indigenous students in Canada, uti-

lizing a database and logistic regression model, shows

promise but has several limitations. Initially, the ap-

proach was highly data-driven, relying on the quan-

tity and availability of data, posing challenges due to

privacy concerns, historical factors, and limited data

collection in Indigenous communities. Missing val-

ues and biased data, along with outliers, can impact

the accuracy of predictions, while sampling bias and

data heterogeneity may hinder model generalization.

Additionally, the logistic regression model, though

insightful, may not capture all complexities, assum-

ing linear relationships and struggling with high-

dimensional and non-numeric data.

Understanding the rich tapestry of diversity

among Canada’s over 600 Indigenous communities

is essential for addressing educational disparities ef-

fectively. Each community, comprising First Na-

tions, M

etis, and Inuit, possesses distinct cultural

and socio-economic backgrounds. Factors such as

cultural diversity, socioeconomic variations, histori-

cal traumas like colonization, and the intertwining of

language and identity directly inﬂuence educational

experiences. Disparities between remote and urban

communities, driven by geographic barriers and in-

frastructure challenges, further complicate the edu-

cational landscape. Limited funding, technical con-

straints, and resistance to technological changes add

to the challenges, requiring collaborative efforts and

strategic initiatives to overcome resource constraints.

While the logistic regression model provides in-

sights into factors contributing to dropout rates, it may

not capture the full spectrum of complexities in In-

digenous education. Its limitations in handling non-

linear relationships, high-dimensional data, and non-

numeric variables highlight the need for a more com-

prehensive approach. Incorporating various machine

learning models such as SVM, decision trees, and sta-

tistical techniques like correlation analysis can offer a

broader understanding of educational disparities. Up-

Addressing Educational Disparities: Assessing the Gap for Indigenous Community

459

holding ethical principles in Indigenous education re-

search, including informed consent, conﬁdentiality,

cultural sensitivity, and community involvement, is

crucial for ensuring positive, inclusive outcomes and

promoting long-term sustainability.

7 CONCLUSION AND FUTURE

WORK

The future work of this research paper holds promis-

ing directions to enhance its impact and tackle ad-

ditional challenges in Indigenous education. Ensur-

ing data privacy and security while collaborating with

Indigenous communities is essential for comprehen-

sive data collection. Cultural sensitivity should be

prioritized throughout the research process to respect

community values and align research outcomes with

their needs. Incorporating qualitative data alongside

quantitative measures can provide deeper insights into

Indigenous students’ experiences. Exploring various

machine learning models and forming collaborations

between academic institutions, government agencies,

and Indigenous communities can lead to more accu-

rate predictions and sustainable change. Ultimately,

future research should focus on promoting equitable

and inclusive education while addressing educational

disparities faced by Indigenous students. In con-

clusion, the logistic regression model proved to be

a valuable tool in addressing educational disparities

faced by Indigenous students. Its accuracy and in-

terpretability allowed for the identiﬁcation of sig-

niﬁcant predictors of dropout, enabling the develop-

ment of evidence-based strategies for educational im-

provement. By incorporating Indigenous data gover-

nance principles and utilizing MongoDB as a NoSQL

database, our management system offers a unique and

comprehensive solution to empower Indigenous com-

munities and promote educational equity in Canada.

The integration of Indigenous knowledge systems,

languages, and histories into the curriculum further

enhances the educational experience for all students

and fosters reconciliation. With our research and

data-driven approach, we aim to contribute to the de-

velopment of effective policies and strategies that pro-

mote justice, fairness, and inclusivity in education for

Indigenous communities.

REFERENCES

Bisong, E. (2019). Logistic regression. pages 243–250.

Brown, D. (2023). Ontario has ‘come quite far’ on indige-

nous education but there’s much more to be done: re-

port. CBC.

Campbell, C. (2020). Educational equity in canada: the

case of ontario’s strategies and actions to advance ex-

cellence and equity for students. School Leadership &

Management, 41(4-5):1–20.

CFS Ontario (2021). Post secondary education and treaty

rights.

Das, A. (2021). Logistic regression. pages 1–2.

Essampally, D. (2020). Logistic regression. Medium.

Government of Canada, Interagency Advisory Panel on Re-

search Ethics (2019). Tri-council policy statement:

Ethical conduct for research involving humans - tcps

2 (2018) - chapter 9: Research involving the ﬁrst na-

tions, inuit and m

etis peoples of canada.

Government of Ontario (2017). Aboriginal students - col-

leges and universities.

K, G. M. (2020). Machine learning basics: Logistic regres-

sion.

Kim, P. J. (2019). Social determinants of health inequities

in indigenous canadians through a life course ap-

proach to colonialism and the residential school sys-

tem. 3(1):378–381.

Li, J., Brar, A., and Roihan, N. (2021). The use of digi-

tal technology to enhance language and literacy skills

for indigenous people: A systematic literature review.

Computers and Education Open, 2.

Misra and Yadav (2020). Improving the classiﬁcation ac-

curacy using recursive feature elimination with cross-

validation. 11(3).

Parker, P. D., Bodkin-Andrews, G., Marsh, H. W., Jerrim,

J., and Schoon, I. (2013). Will closing the achieve-

ment gap solve the problem? an analysis of primary

and secondary effects for indigenous university entry.

Journal of Sociology.

Rao, K. M., Saikrishna, G., and Supriya, K. (2023). Data

preprocessing techniques: emergence and selection

towards machine learning models - a practical review

using hpa dataset. Multimedia Tools and Applications.

Schaub, A. (2020). Examining an apparent educating gap

between non-indigenous and indigenous learners: A

hermeneutic phenomenology approach.

Stagg Peterson, S. and Dwyer, B. (2016). Research in

canada’s northern rural and indigenous communities:

Supporting young children’s oral language and writ-

ing. The Reading Teacher, 70.

Weston, D. (2019). Importance of safe water for aboriginal

children’s education. Heart and Art.

Wu, J., Lin, S., Kong, H., and Shi, H. (2019). The com-

bination forecasting model of telecommunication user

tricking account overdraft limit based on logistic re-

gression and svm. pages 411–415.

CSEDU 2024 - 16th International Conference on Computer Supported Education

460