EREO: An Effective Rule Evaluation Framework for Discovering

Interesting Patterns in US Birth Data and Beyond

Abhilash C. B.

1 a

and Kavi Mahesh

2 b

Indian Institute of Information Technology Dharwad, IIIT Dharwad, India

Department of Data Science and Intelligent Systems, Indian Institute of Information Technology Dharwad,

Keywords:

Ontology, Association Rule Mining, Rule Evaluation, Interestingness Measures, Integrated Rule Information

Content (IRIC) US Birth Data.

Abstract:

Birth data holds immense importance in healthcare for several reasons. It offers a comprehensive and rep-

resentative sample of the population, enabling the identiﬁcation of patterns and trends that can signiﬁcantly

impact public health policies and interventions. However, extracting interesting patterns from the vast birth

data attributes poses a domain-speciﬁc and challenging problem. We can derive intriguing patterns by utilizing

rare rules for identifying interesting associations. The level of interestingness depends on various factors, in-

cluding the user, data, and domain. To address this, we propose the Effective Rule Evaluation using Ontology

(EREO) framework, which incorporates two modes of rule evaluation. Firstly, the Integrated Rule Information

Content (IRIC) measure is employed to quantify the level of interestingness. Secondly, the interesting rules

are assessed by domain experts. The combined approach of these two modes of evaluation conﬁrms the level

of interestingness of the derived rules.The study demonstrates a signiﬁcant relationship between these two

modes of assessment, providing evidence of the convergence between expert evaluations and the ontology-

based association rule measurements. This connection adds further value to the ﬁeld by contributing to the

understanding and measurement of interestingness within the context of ontology-based association rules

1 INTRODUCTION

Birth data is a valuable resource in healthcare research

and analysis, providing a comprehensive and repre-

sentative sample of the population (Abhilash and Ma-

hesh, 2022). The patterns and trends identiﬁed in

birth data can have signiﬁcant implications for public

health policies and interventions. However, extracting

interesting patterns from the vast array of birth data

attributes presents a challenging and domain-speciﬁc

problem (C and Mahesh, 2021; Zhou et al., 2020;

Tandan et al., 2021).

In this study, our focus is on addressing this chal-

lenge by leveraging rare rules to derive interesting as-

sociations and patterns from birth data (Abhilash and

Mahesh, 2023). The concept of interestingness is sub-

jective and depends on factors such as user prefer-

ences, data characteristics, and the speciﬁc domain of

investigation. To address this, we propose an Effec-

tive Rule Evaluation using Ontology (EREO) frame-

https://orcid.org/0000-0002-4864-8485

https://orcid.org/0000-0002-4353-725X

work, which utilizes two modes of rule evaluation

(Abhilash and Mahesh, 2023).

The ﬁrst mode of evaluation involves the In-

tegrated Rule Information Content (IRIC) measure

(Manda et al., 2015), which quantiﬁes the level of

interestingness of the derived rules. By considering

the information content embedded within the rules,

IRIC helps identify the most intriguing and relevant

patterns. The second mode of evaluation incorporates

domain experts’ assessment of interesting rules. Their

expertise and insights play a vital role in conﬁrming

and validating the interestingness level of the derived

patterns (Manda et al., 2013; Bringmann et al., 2011;

Geng and Hamilton, 2006).

This study aims to demonstrate the efﬁcacy and

reliability of the EREO framework in evaluating

the interestingness of rules derived from birth data.

By employing both quantitative measures and expert

evaluation, we aim to provide a systematic approach

to identifying and interpreting interesting patterns.

Through extensive experimentation and evalua-

tion, we establish a signiﬁcant relationship between

568

C. B., A. and Mahesh, K.

EREO: An Effective Rule Evaluation Framework for Discovering Interesting Patterns in US Birth Data and Beyond.

DOI: 10.5220/0012137400003541

In Proceedings of the 12th International Conference on Data Science, Technology and Applications (DATA 2023), pages 568-575

ISBN: 978-989-758-664-4; ISSN: 2184-285X

 2023 by SCITEPRESS – Science and Technology Publications, Lda. Under CC license (CC BY-NC-ND 4.0)

the two rule assessment modes, conﬁrming the pro-

posed framework’s effectiveness. The results of this

study contribute to the ﬁeld of healthcare analytics by

enhancing our understanding of patterns and associa-

tions within birth data and providing valuable insights

for public health decision-making.

Our contributions are as follows:

• We propose an Effective Rule Evaluation using

Ontology (EREO) framework for evaluating the

level of interestingness.

• We introduce an effective measure called Inte-

grated Rule Information Content using Ontology

(IRIC) for calculating the interestingness score.

Furthermore, the interesting rules are assessed by do-

main experts to ensure correctness, taking into con-

sideration the parameter of clinical relevance.

Overall, this research highlights the importance of

birth data analysis and proposes a novel framework,

EREO, for evaluating the interestingness of associa-

tion rules. By leveraging rare rules and incorporating

expert evaluation, the framework enables the identi-

ﬁcation of meaningful patterns in birth data, offering

potential beneﬁts for public health policies, interven-

tions, and improved healthcare outcomes.

The remainder of the paper is structured as fol-

lows: Section 2 discusses the recent literature. Sec-

tion 3 presents the data and methodology for deriving

interesting inferences from US birth data. Section 4

showcases the results along with the interestingness

scores. Finally, Section 5 concludes the paper and

highlights future research directions.

2 LITERATURE REVIEW

The literature surrounding the analysis of birth data

and the evaluation of interesting patterns provides

valuable insights into the existing research and

methodologies in this ﬁeld.

One area of focus in the literature is the ex-

ploration of association rule mining techniques for

discovering patterns in birth data. Various studies

have employed association rule mining algorithms to

identify associations between maternal characteris-

tics, birth outcomes, and demographic factors. These

studies have shed light on risk factors for adverse

birth outcomes, such as preterm birth and low birth

weight, as well as the impact of socioeconomic fac-

tors on birth outcomes (Bekkar et al., 2020; Griggs

et al., 2020; Robati et al., 2020).

Additionally, researchers have proposed different

measures of interestingness to quantify the signiﬁ-

cance and relevance of discovered patterns. Measures

like support, conﬁdence, and lift have been commonly

used to assess the interestingness of association rules

in birth data. These measures aid in prioritizing and

selecting the most meaningful and relevant patterns

for further analysis and interpretation.

In recent years, the integration of ontologies in

rule evaluation has gained attention in the literature.

Ontologies provide a structured and standardized rep-

resentation of domain knowledge, enabling the incor-

poration of domain expertise into the evaluation pro-

cess. OntoVPA is a commercially available Dialogue

Management System for Virtual Personal Assistants

(VPAs) that utilizes ontologies and ontology-based

rules (Wessel et al., 2019).

The approach for curating Gene Ontology (GO)

data by utilizing association rule mining and intro-

ducing a speciﬁcity measure called VICD. The results

demonstrate that VICD scores have stronger correla-

tions with speciﬁcity values and lead to more consis-

tent association rules (Shui and Cho, 2016).

(Mattiev and Kavsek, 2020), This study presents

a new method for constructing accurate and compact

classiﬁers by reducing the number of class associa-

tion rules. The proposed associative classiﬁer selects

strong rules based on overall coverage and achieves

high classiﬁcation accuracy while generating smaller

rules compared to traditional classiﬁers.

However, there remains a need for an effective

framework that combines robust rule evaluation tech-

niques, such as ontology-based measures, with ex-

pert evaluation to ensure the correctness and clinical

relevance of the interesting rules. This gap in the

literature highlights the importance of developing a

comprehensive approach for evaluating the interest-

ingness of association rules derived from birth data.

To address this gap, our study proposes the Effec-

tive Rule Evaluation using Ontology (EREO) frame-

work, which leverages the Integrated Rule Informa-

tion Content (IRIC) measure and domain expert eval-

uation. The IRIC measure accounts for the informa-

tion content embedded within the rules, providing a

quantitative indicator of interestingness. The involve-

ment of domain experts ensures the assessment of in-

teresting rules in terms of their clinical relevance and

alignment with current healthcare practices.

By reviewing the existing literature, we have iden-

tiﬁed the need for an integrated approach that com-

bines ontology-based measures, expert evaluation,

and the domain-speciﬁc nature of interestingness as-

sessment in the context of birth data analysis. Our

study aims to contribute to this ﬁeld by introducing

the EREO framework and showcasing its effective-

ness in evaluating the interestingness of association

rules derived from US birth data.

EREO: An Effective Rule Evaluation Framework for Discovering Interesting Patterns in US Birth Data and Beyond

569

3 DATA AND METHODS

This section provides a comprehensive overview of

our research’s data, ontology, and methodology.

3.1 Data and Ontology

In this study, we used the US birth data from 2020

as our primary dataset. The details of this dataset are

presented in Table 1.

Table 1: Dataset and Ontology Information.

Dataset: Birth Data 2020

Access Link:

Instances: 1 Million

Ontology: BirthOnto

Link:

The birth data attributes encompass a wide range

of information related to maternal characteristics,

birth outcomes, demographic factors, and other rele-

vant variables. These attributes include maternal age,

race/ethnicity, education level, gestational age, birth

weight, and geographical location. The inclusion of

these attributes allows for a comprehensive analysis

of various factors that may inﬂuence birth outcomes

and patterns. To facilitate the evaluation of interest-

ingness and enhance the interpretation of the derived

association rules, we incorporated ontology into our

methodology. Ontology provides a structured rep-

resentation of domain knowledge and allows for the

integration of expert-deﬁned concepts and relation-

ships. By incorporating ontology, we introduce a

standardized framework that enables more meaning-

ful and contextually relevant rule evaluation.

To illustrate the importance of ontology, consider

an example where the birth data attributes include ma-

ternal education level and birth weight. By utilizing

ontology, we can deﬁne and establish relationships

between these attributes and relevant concepts, such

as ”high-risk pregnancy” or ”low birth weight risk

factors.” This enables a more nuanced and detailed

analysis of the association rules, leading to valuable

insights into the relationships between maternal edu-

cation level, birth weight, and their impact on birth

outcomes.

3.2 Methodology

3.2.1 Hypothesis for Rare Rules

• Infants born to mothers who smoked cigarettes

before pregnancy and had pre-pregnancy hyper-

tension are more likely to be born preterm.

Figure 1: EREO Framework.

• Infants born to mothers aged more than 35 and

had Hypertension Eclampsia have a higher likeli-

hood of being born preterm.

• Infants born to mothers who had gonorrhoea and

chlamydia during pregnancy have a higher likeli-

hood of being admitted to intensive care.

• Infants born to mothers aged more than 35 who

smoke have a higher likelihood of having Down

syndrome.

• Mothers who had Failed External Cephalic Ver-

sion and had Gestational Diabetes resulted in in-

fants delivered through cesarean delivery.

• Mothers with a history of smoking and hyperten-

sion are at higher risk of developing cardiovascu-

lar diseases.

• Mothers who smoke during pregnancy are more

likely to deliver infants with low birth weight than

mothers who do not smoke during pregnancy.

3.2.2 Integrated Rule Information Content

(IRIC)

IRIC (Integrated Rule Information Content) is a mea-

sure used to evaluate the interestingness of associa-

tion rules in ontology-based data mining. It considers

the information content of the predictor and outcome

concepts in a rule, as well as the shared information

between them. By calculating and combining the in-

formation content with weights, IRIC provides a com-

prehensive measure of the rule’s interestingness. It is

computed as shown in equation 1.

IRIC (X → Y ) = ((α ∗N

(X))+ (β ∗ N

(Y ))) ∗ N

COMI

(X → Y )

(1)

Where:

• N

(X) is the information content of concept X

• N

(Y ) is the information content of concept Y

• N

COMI

(X → Y ) is the shared information between

concepts X and Y

• α and β are weights assigned to concepts X and Y

respectively, with α + β = 1

DATA 2023 - 12th International Conference on Data Science, Technology and Applications

570

Figure 2: Rules Generated for Hypothesis.

To calculate N

, you can use the formula:

(t) = −log



p(t)

UB(IC)



(2)

Where:

• t is a concept in the ontology

• p(t) is the probability of t, calculated as the num-

ber of instances in the data that are annotated with

t or any of its descendants, divided by the total

number of instances

• U B(IC) is the upper bound for information con-

tent, calculated as −log





, where N is the total

number of instances in the data

Once you have calculated N

and N

COMI

for each

rule, you can use IRIC to rank the rules in order of

interestingness.

4 RESULTS AND DISCUSSION

The hypotheses deﬁned in our study have been gener-

ated through collaborative consultations with doctors

aimed at exploring speciﬁc relationships and patterns

within the US Birth Data of 2020. These hypotheses

provide valuable insights into potential associations

between various factors and birth outcomes. The ta-

ble below presents the hypotheses, along with their

corresponding statuses. Table 2 represents the deﬁned

hypothesis status.

Figure 2 represent the number of interesting

rules that were generated to support the hypothesis.

The Accept hypotheses uncover signiﬁcant correla-

tions among various maternal characteristics, such

as smoking, age, pre-pregnancy hypertension, gesta-

tional infections, and speciﬁc birth outcomes. These

associations provide valuable insights into identify-

ing risk factors and offer potential avenues for imple-

menting interventions to address issues like preterm

birth, intensive care admissions, and Down syndrome.

Table 2: Hypothesis from US Birth Data 2020.

# Hypothesis Status

H1 Infants born to mothers who

smoked cigarettes before preg-

nancy and had pre-pregnancy

hypertension are more likely to be

born preterm.

H2 Infants born to mothers aged more

than 35 and had Hypertension

Eclampsia have a higher likelihood

of being born preterm.

H3 Infants born to mothers who had

gonorrhoea and chlamydia during

pregnancy have a higher likelihood

of being admitted to intensive care.

H4 Infants born to mothers over 35 who

smoke have a higher likelihood of

Down syndrome.

H5 Mothers who had Failed External

Cephalic Version and had Gesta-

tional Diabetes resulted in infants

delivered through cesarean delivery.

H6 Mothers with a history of smoking

and hypertension are at higher risk

of developing cardiovascular dis-

eases.

Reject

H7 Mothers who smoke during preg-

nancy are more likely to deliver in-

fants with low birth weight than

mothers who do not smoke during

pregnancy.

Reject

4.1 Semantic Interesting Rules

In the context of our work, the term ”Semantic In-

teresting Rules” refers to association rules that incor-

porate the semantic knowledge captured by the on-

tology. In this study we have used Birthonto

. The

results are represented in Table 3 4, and 5, the “Pre-

dictor 0” and “Predictor 1” columns indicate the an-

tecedent part of the rule. The “Outcome” column

speciﬁes the consequent part of a rule. The “IRIC”

column denotes the Integrated Rule Information Con-

tent, and “IS” column represents the Interestingness

Score, indicating the degree of Interestingness of the

rule. The “IRIC” values assigned to each rule in the

table represent the Integrated Rule Information Con-

tent, which measures the level of interestingness of

the rule. A higher IRIC value suggests a stronger as-

https://bioportal.bioontology.org/

EREO: An Effective Rule Evaluation Framework for Discovering Interesting Patterns in US Birth Data and Beyond

571

sociation between the predictor variables and the out-

come. Additionally, the “‘IS” column indicates the

interestingness score of each rule, reﬂecting the de-

gree of interest in the association. The IS is computed

considering the IRIC values with the deﬁned thresh-

old.

The clinical signiﬁcance of Table 3 resides in its

depiction of the correlations between predictor vari-

ables and the occurrence of Neonatal Intensive Care

Unit (NICU) admissions. The table speciﬁcally in-

vestigates the relationship between the predictor vari-

ables “CigarettesBeforePregnancyRecode” and “Pre-

pregnancyHypertension” and their impact on NICU

admission outcomes. Each row in the table represents

a distinct scenario or range of “CigarettesBeforePreg-

nancyRecode” values alongside their corresponding

associations with NICU admissions, as outlined in the

“Outcome” column.

Table 4 examines the relationships between mater-

nal age and the presence of hypertension eclampsia in

relation to the outcome of admission to the Neonatal

Intensive Care Unit (NICU). It explores different age

ranges (35-39 years, 50-54 years, 40-44 years, and

45-49 years) and their associations with NICU admis-

sion.

Table 5 represents the interesting rules generated

based on the hypotheses H3, H4, and H5. The asso-

ciation between gonorrhoea and chlamydia with ad-

mission to the intensive care unit, smoking status and

maternal age with conﬁrmed Down syndrome, and

failed external cephalic version and gestational di-

abetes with the cesarean delivery method are high-

lighted.

The distribution of IRIC values from the three ta-

bles reveals the relative interestingness of the rules.

The rules with higher IRIC values are associated with

lower distribution values, indicating higher interest-

ingness.

Overall, utilizing the semantic information encap-

sulated within the ontology, we can uncover valu-

able insights and identify associations that might be

overlooked by conventional rule-mining techniques.

The semantic interesting rules provide a heightened

understanding of the interrelationships between di-

verse attributes and outcomes within the birth data,

enabling more informed decision-making in the realm

of healthcare.

Figure 3 represents the boxplot that displays the

distribution of the IRIC values derived from the data.

By observing the boxplot, you can gain insights into

the dispersion and distribution characteristics of the

IRIC values.

Referring to Figure 4, the distribution of IRIC val-

ues from the three tables reveals the relative interest-

Figure 3: Box Plot of IRIC Values.

Figure 4: Distribution of IRIC Values.

ingness of the rules. The rules with higher IRIC val-

ues are associated with lower distribution values, in-

dicating higher interestingness.

4.2 Domain Experts Evaluation

In our study, we recognized the importance of incor-

porating domain expert analysis to validate and eval-

uate the generated rules and their interestingness. We

sought the expertise of doctors who served as domain

experts in the ﬁeld of healthcare and maternal care.

Their insights and evaluations were invaluable in as-

sessing the relevance and signiﬁcance of the rules de-

rived from the US Birth Data.

We employed an Interestingness Measurement

Scale (IMS) proposed by (CB et al., 2023; Abhilash

and Mahesh, 2023) to facilitate the evaluation pro-

cess. The evaluation conducted by the domain ex-

perts revealed a signiﬁcant positive relation between

the Integrated Rule Information Content (IRIC) val-

ues and their expert evaluation. The evaluation was

taken using google forms. Figure 5 represents the do-

main experts evaluation of all ﬁve hypothesis. Table

6 indicates the normalised IRIC scores with the aver-

age domain expert evaluation score of the interesting

rules.

DATA 2023 - 12th International Conference on Data Science, Technology and Applications

572

Table 3: Interesting Rule for Hypothesis-1.

Predictor 0 Predictor 1 Outcome IRIC IS

CigarettesBeforePregnancyRecode

: 21-40

Pre-pregnancyHypertension :

Yes

AdmissionToNICU :

Yes

23.41 1

CigarettesBeforePregnancyRecode

: 1-5

Pre-pregnancyHypertension :

Yes

AdmissionToNICU :

Yes

21.41 1

CigarettesBeforePregnancyRecode

: 41 or more

Pre-pregnancyHypertension :

Yes

AdmissionToNICU :

Yes

24.26 1

CigarettesBeforePregnancyRecode

: 6-10

Pre-pregnancyHypertension :

Yes

AdmissionToNICU :

Yes

19.15 1

CigarettesBeforePregnancyRecode

: 11-20

Pre-pregnancyHypertension :

Yes

AdmissionToNICU :

Yes

18.33 1

Table 4: Interesting Rule for Hypothesis-2.

Predictor 0 Predictor 1 Outcome IRIC IS

MotherAge : 35-39 Years HypertensionEclampsia :

Yes

AdmissionToNICU :

Yes

17.31 1

MotherAge : 50-54 Years HypertensionEclampsia :

Yes

AdmissionToNICU :

Yes

38.39 1

MotherAge : 40-44 Years HypertensionEclampsia :

Yes

AdmissionToNICU :

Yes

18.95 1

MotherAge : 45-49 Years HypertensionEclampsia :

Yes

AdmissionToNICU :

Yes

7.03 0

Table 5: Interesting Rule for Hypothesis-3,4 and 5.

# Predictor 0 Predictor 1 Outcome IRIC IS

H3 Gonorrhea : Yes Chlamydia : Yes AdmitToIntensiveCare : Yes 46.78 1

H4 SmokeStatus : Smoker MotherAge : 40-44

Years

DownSyndrome : Con-

ﬁrmed

49.41 1

H4 SmokeStatus : Smoker MotherAge : 35-39

Years

DownSyndrome : Con-

ﬁrmed

8.51 0

H5 FailedExternalCephalicVersion :

Yes

GestationalDiabetes :

Yes

FinalMethodOfDelivery :

Cesarean

18.18 1

1. IRIC: Integrated Rule Inﬁrmation Content.

2. IS: Interestingness Score.

3. Predictor: Antecedent of rule.

4. Outcome: Consequent of Rule.

Figure 6 indicates the comparison between the

two levels of EREO framework. The RMS value of

0.534 suggests that there is some difference or dis-

crepancy between the domain experts’ scores and the

normalized hypothesis scores. However, it is impor-

tant to note that the magnitude of this difference is

relatively small, indicating a moderate level of agree-

ment between the two sets of scores. Also, as a statis-

tical evaluation, we used T-test for the values in Ta-

ble 6, it found that the p-value of 0.3103 indicates

that there is a 31.03% probability of obtaining the ob-

served difference in means by chance alone, assuming

that there is no true difference between the two sets of

scores.

EREO: An Effective Rule Evaluation Framework for Discovering Interesting Patterns in US Birth Data and Beyond

573

Figure 5: Domain Experts Evaluation of Rules.

Table 6: Comparison of Domain Experts Scores and Nor-

malized Hypothesis Scores.

Hypotheses Domain Experts

Score

Normalized

IRIC Score

H1 3.9 3.60

H2 3.8 3.45

H3 4.0 5.0

H4 4.2 4.03

H5 3.3 2.85

Figure 6: Comparision of IRIC and Domain Experts Evalu-

ation.

4.3 Future Research Directions

Future research directions in this ﬁeld involve the in-

corporation of additional measures for evaluating in-

terestingness that align with the ontology-based as-

sociation rule framework. By introducing AI-based

evaluation as a third mode within the EREO frame-

work, researchers can explore and leverage advanced

techniques to further enhance the assessment of inter-

estingness.

5 CONCLUSIONS

In summary, our study examined the relationships

between various maternal characteristics and spe-

ciﬁc birth outcomes using US Birth Data from 2020.

Through collaborative consultations with medical ex-

perts, we formulated hypotheses that provided valu-

able insights into these associations. By leveraging

the semantic information embedded in the birth data

ontology, we discovered deeper insights and uncov-

ered connections that traditional rule mining tech-

niques may overlook. The semantic interesting rules

we generated enhanced our understanding of how dif-

ferent attributes relate to birth outcomes. These rules

revealed important risk factors and potential inter-

ventions for preterm birth, intensive care admissions,

Down syndrome, and other relevant outcomes. The

clinical signiﬁcance of these ﬁndings lies in their abil-

ity to guide healthcare decision-making and facilitate

targeted interventions.

The insights gained from this study have impli-

cations for improving healthcare practices, identify-

ing at-risk populations, and implementing preven-

tive measures. Further exploration and analysis in

this area can advance our knowledge and support

evidence-based decision-making in the healthcare do-

main.

ACKNOWLEDGEMENTS

This work is supported in part by the Indian Institute

of Information Technology Dharwad (IIIT Dharwad).

We would like to express our gratitude to the two

anonymous reviewers who have contributed to the im-

provement of the current version of the paper. We also

thank JSS Mahavidyapeetha and JSSATE, Bangalore,

for their continuous support and encouragement.

REFERENCES

Abhilash, C. and Mahesh, K. (2022). Ontology-based inter-

estingness in covid-19 data. In Metadata and Seman-

tic Research: 15th International Conference, MTSR

2021, Virtual Event, November 29–December 3, 2021,

Revised Selected Papers, pages 322–335. Springer.

Abhilash, C. and Mahesh, K. (2023). Ontology is what

makes data interesting: Interestingness framework for

covid-19 corpora. Journal of Information Science.

Bekkar, B., Pacheco, S., Basu, R., and DeNicola, N.

(2020). Association of air pollution and heat expo-

sure with preterm birth, low birth weight, and stillbirth

in the us: a systematic review. JAMA network open,

3(6):e208243–e208243.

Bringmann, B., Nijssen, S., and Zimmermann, A. (2011).

Pattern-based classiﬁcation: A unifying perspective.

arXiv preprint arXiv:1111.6191.

C, A. and Mahesh, K. (2021). Graph analytics applied to

covid19 karnataka state dataset. In 2021 The 4th Inter-

DATA 2023 - 12th International Conference on Data Science, Technology and Applications

574

national Conference on Information Science and Sys-

tems, page 74–80, New York, NY, USA. Association

for Computing Machinery.

CB, A., Mahesh, K., and Sanda, N. (2023). Ontology-based

semantic data interestingness using bert models. Con-

nection Science, 35(1):2190499.

Geng, L. and Hamilton, H. J. (2006). Interestingness mea-

sures for data mining: A survey. ACM Computing

Surveys (CSUR), 38(3):9–es.

Griggs, K. M., Hrelic, D. A., Williams, N., McEwen-

Campbell, M., and Cypher, R. (2020). Preterm la-

bor and birth: a clinical review. MCN: The American

Journal of Maternal/Child Nursing, 45(6):328–337.

Manda, P., McCarthy, F., and Bridges, S. M. (2013). Inter-

estingness measures and strategies for mining multi-

ontology multi-level association rules from gene on-

tology annotations for the discovery of new go

relationships. Journal of biomedical informatics,

46(5):849–856.

Manda, P., McCarthy, F., Nanduri, B., Wang, H., and

Bridges, S. M. (2015). Information-theoretic inter-

estingness measures for cross-ontology data mining.

arXiv preprint arXiv:1504.08027.

Mattiev, J. and Kavsek, B. (2020). Coverage-based clas-

siﬁcation using association rule mining. Applied Sci-

ences, 10(20):7013.

Robati, S., Saab, W., Dur

an-Retamal, M., Saab, W.,

Theodorou, E., Cawood, S., Serhal, P., and Seshadri,

S. (2020). The association between elevated proges-

terone level on day of hcg trigger and live birth rates in

art cycles: a single centre observational study. Journal

of Reproduction & Infertility, 21(4):283.

Shui, Y. and Cho, Y.-R. (2016). Filtering association rules

in gene ontology based on term speciﬁcity. In 2016

ieee international conference on bioinformatics and

biomedicine (bibm), pages 1314–1321. IEEE.

Tandan, M., Acharya, Y., Pokharel, S., and Timilsina, M.

(2021). Discovering symptom patterns of covid-19

patients using association rule mining. Computers in

biology and medicine, 131:104249.

Wessel, M., Acharya, G., Carpenter, J., and Yin, M.

(2019). Ontovpa—an ontology-based dialogue man-

agement system for virtual personal assistants. In

Advanced Social Interaction with Agents: 8th Inter-

national Workshop on Spoken Dialog Systems, pages

219–233. Springer.

Zhou, F., Yu, T., Du, R., Fan, G., Liu, Y., Liu, Z., Xiang,

J., Wang, Y., Song, B., Gu, X., et al. (2020). Clinical

course and risk factors for mortality of adult inpatients

with covid-19 in wuhan, china: a retrospective cohort

study. The lancet, 395(10229):1054–1062.

EREO: An Effective Rule Evaluation Framework for Discovering Interesting Patterns in US Birth Data and Beyond

575