1 Introduction

For many reasons, understanding the factors that influence children’s academic skills is imperative. Research findings show that children’s cognitive and non-cognitive development influence their achievements and decisions [1].

In each generation, people’s behavior changes with the development of technology. Children at present utilize more and more time with technological gadgets than in the past; this is true even for infants [2]. In the literature, some studies investigate the relationship between the amount of time spent watching television, the amount of sleeping time, a child’s physical fitness, weight, and academic achievement [3, 4]. Anderson [5] revealed that 26% of all US children watch four or more television hours daily and that 67% watch at least two hours daily. Gortmaker [6] found that for children between the ages of 10 and 15, television viewing is a significant contributing factor for a child to be overweight. In addition, the research indicates that reducing viewing time can keep children from developing some common chronic health conditions.

Body Mass Index (BMI) is an indicator that identifies the level of obesity. Studies conducted by Trembly [7], Wingfield [8], and Cawley [9] have studied the relationship between BMI and physical fitness. Study conducted by Cawley [9] found that physical exercise lowers the BMI z-score and reduces obesity among 5th graders, especially boys. While [7] and [8] found the existence of a negative relationship between BMI and physical activity, the findings of [8] indicate that the negative relationship between BMI and physical fitness is significant across grade levels and gender. Wingfield [8] Datar [10], Taras and Potts-Datema [11], Castelli [12], and Trembly [7] considered the impact of weight and physical fitness on academic skills. Both [10] and [11] found a negative association between academic skills and obesity in school-aged children. Specifically, [10] found that the performance in mathematics and reading scores of overweight children is lower than that of non-overweight children in kindergarten. In a study using 259 public school third and fourth graders, [12] found an inverse relationship between pre-adolescents BMI and academic skills and a positive correlation between physical fitness and academic skills. Wingfield [11] revealed a correlation between the academic skills of 5th-grade females and BMI and physical fitness. In another study, [7] examined the relationships between 6th-grade students’ physical activity level, BMI, self-esteem, reading scores, and mathematics scores. In the same study, physical activity and self-esteem were positively related, and there was a trivial negative relationship between physical activity and academic achievement. Using a longitudinal study, Benner and Crosnoe [13] show a strong association between children’s academic and school racial/ethnic composition. More studies on students’ academics and race can be found [7,8,9,10,11,12,13,14]. Another study conducted by Stroebele [15] using self-reported students’ related factors of elementary students, overweight, less physical activity, food consumption habits, and the number of hours of television watching negatively impact students' academic success. Furthermore, students who slept at least 9 h showed higher academic performances.

According to [16] and [17], students’ academic performance is influenced by the motivational effects of the family, peers, and the classroom atmosphere. The influence of peers can negatively impact other students. As Ogbu [18] states, at times, some groups of students tend to interpret their academic failures as accomplishments rather than their failures. Another study conducted by Johnson [19] shows that peer effect strongly influences elementary students’ academic achievements, while family background is also an influential factor. In addition, [19] states that girls’ academic performance is better at reading and writing, while boys perform better than their counterparts in more analytical subject disciplines.

Furthermore, [19] states substantial differences in academic achievement among students of different races, though this difference narrows down as age progresses.

One of the educational practices is to group students according to their educational level. The practice of grouping students emerged as students' preparation for their future workforce [20]. Grouping students based on their education level increases their academic achievements as the instructors have the freedom to adjust the pace of the instructions to the group of nearly uniform students. Furthermore, a more challenging and higher-level curriculum can be implemented for a higher level of kids. According to [21] elementary kids' grouping becomes maximally effective when applied only to one or two subjects. Some researchers find adverse effects of within-class ability grouping. Therefore, classifying students based on their education level is valuable for educators and policymakers. A comprehensive study about grouping practices is conducted by [22].

Grouping students or classifying them based on educational performance was benefitted by the development of machine-learning techniques. In this direction, [23] utilizes various machine-learning techniques to develop a model to classify students based on their academic levels. In another study, [24] use some of the demographic characteristics of the students together with their written assignment grades to predict students’ performance in a distance learning class. Ahadi [24] conducted a study to investigate how to use machine-learning techniques to identify students who are underperforming and high-performing. In another study, Peterson [25] used machine learning algorithms to assess students’ reading levels. More applications of machine-learning techniques related to students’ academic performances can be found in [27] and [28]. Though there are numerous studies related to student’s academic performance using machine-learning techniques at various academic levels, there is a vacuum of such studies related to elementary students.

As pointed out before, it is essential to find out the relationship among the factors that influence the academic success of elementary students. Furthermore, gathering early information about the student’s educational level is helpful for many viewpoints, including the parents, instructors, and administrators.

Studying this direction, we answer this manuscript’s research questions: (1) Which variables impact first and second-graders academic success? (2) How can we predict students' academic success at an early stage? (3) How can students be classified into appropriate academic groups based on student-related variables? Therefore, the outcome of a study like this will be helpful to the parents, instructors, administrators, and literature.

2 Materials and methods

2.1 Sample

This dataset contains 185 students, including 91 first and 94 s graders. Out of the first graders, 51% are boys, and 49% are girls, while for the second graders, these percentages are 56 and 44%, respectively. As far as the racial percentages are concerned, out of the first graders, 66% of the students were white, 13% black, and the rest were from other groups, while out of the second graders, 69% were white, 24% were black, and the rest were from other groups. For this study, we recorded the following variables: the student's grade, gender, race, weight, height, the number of hours students sleep, involvement in physical activities, and reading books.

2.2 Study design

This study used a cross-sectional quantitative research methodology [29]. Data for this investigation was implemented at an elementary school in rural New Mexico, USA. The school has classes from grade 1 to grade 2, starting at 8:00 am and ending at 3:30 pm. First and second-graders curriculum consists of disciplines including Mathematics, Language and Arts, Social Studies, Science, and Physical Education.

The data for this study was collected with the approval of the students, parents, and the school. Students who provided consent were considered for this study, while others were discarded. Students, parents, teachers, and the nurses at the school assist with data collection for this study.

3 Measuring instruments

3.1 DIBELS score

Dynamic Indicators of Basic Early Literacy Skills (DIBELS) score is a measurement to evaluate the acquisition of early-stage literacy skills for K-6 grade students. Furthermore, this score measures the students’ early literacy skills development and is monitored periodically at the academic year’s beginning, middle, and end. The DIBELS score reflects students’ phonemic awareness, alphabetic principle, accuracy, fluency, and comprehension. Out of the three DIBELS scores, we use the end-of-the-year DIBELS scores to represent students’ academic performance indicators. Further information about the DIBELS score can be found in [29].

3.2 Data collection using a questionnaire

After obtaining consent from all the parties involved in this study, a questionnaire was developed and distributed to the parents of the students through the school teachers. The questionnaire tool employed consisted of the following personal and Gender: Student’s Gender: Student’s gender was recorded as 0-Female and 1-Male.


Race: Race was recorded as American Indian or Alaskan Native, Asian, Black or African American, Native Hawaiian or Other Pacific Islander, and White. Due to the fewer number of students from certain race groups, the final representation of the race data was recorded using the three categories of white (1), black (2), and other (3).


TV: Each student’s average number of hours spent on watching TV in a day Sleep: Each student’s average hours of sleep in a day. Though this variable has a continues nature, in the later part of the analysis, the variable was categorized as LESS (sleeping time is less than 9 h) and GOOD (sleeping time is between 9 and 12 h) as the research has shown higher academic results with students who make at least 9 h of sleep [15].


Books: Each student’s average number of hours of reading books in a day.


Physical: This includes each student’s duration of doing physical activities such as playing, cycling, running, etc.


Height and Weight: Each student’s height and weight were measured with the help of the school nurse.


BMI: BMI values were calculated using the heights and weights obtained according to the following equation provided by [31].

$${\text{BMI}} = \, \left( {{\text{weight }}\left( {{\text{kg}}} \right)/\left( {{\text{height }}\left( {\text{m}} \right){2}} \right)} \right)$$

4 Data analysis

4.1 Statistical indicators

We calculate descriptive statistics such as measures of central tendency, variability, and measures of the relative position of each variable collected on the students. In addition, we observe the shape of distributions of the critical variables related to this dataset. Some of the descriptive statistics we computed are as follows. Mean, standard deviation, DIBELS scores for both boys and girls are symmetrically distributed centered about 225, though girls’ scores show less variability than the boys. The distributions of BMI values of both boys and girls are skewed to the right. Like the DIBELS scores, BMI values for girls show less variability than for boys. In addition, the distribution of sleeping times also are approximately normally distributed around 8.5 h.

4.2 Two-way ANOVA and post-hoc methods

Two-way ANOVA (Analysis of Variance) is a statistical method used to analyze the effects of two categorical independent factors on a continuous dependent variable. It determines whether there are significant interactions between the two factors and whether each factor has a significant main effect on the dependent variable. In this study, we use two-way ANOVA to identify the impact of gender and sleep duration.

Further analysis is needed to identify the impact of gender and sleep duration on the students’ DIBELS scores. Though the sleeping duration is a continuous variable, here, for the ANOVA analysis, we discretize it by categorizing it into categories, namely LESS and GOOD, as explained before.

The null hypotheses of this test are as follows:

  • The means of observations grouped by gender are the same.

  • The means of observations grouped by the duration of sleeping are the same.

  • There is no interaction between the above two factors.

The alternative hypotheses are that the means of the above groups are not the same and that there is an interaction effect.

In order to make a decision with the ANOVA test, we use the F value and p-values.

As ANOVA does not determine which levels of the main factors are the highest and the lowest, further analysis is required to identify the above. This step is called a post-hoc analysis. We use a post-hoc analysis called Tukey’s HSD to determine which factor levels are significantly different from each other.

4.3 Regression analysis

Multiple linear regression is a statistical technique used to understand the linear association between a dependent variable, which is continuous, and two or more independent variables, which are either discrete or continuous. It extends the concept of simple linear regression, where only one independent variable is used to predict the dependent variable. In this study, we used the DIBELS score as the dependent variable, and the rest of the variables we collected about students, such as gender, the time they spend reading books, sleeping, and doing physical activities, and BMI. Even with the regression analysis, we use p-values to determine the impact of each independent variable.

4.4 Machine learning models

Machine learning, a subset of artificial intelligence, comprises computer systems capable of learning and adapting without predefined instructions. In contrast to the standard statistical procedures, machine learning does not follow parametric assumptions, which enhances the applicability to various data sets from various domains. This study uses two machine learning techniques called Naïve Bayes and Decision Tree algorithms.

4.4.1 Naïve bayes (NB) model

In this study, we use the NB model to group students into their appropriate academic groups based on their characteristics belonging to the students. This classification model is constructed using the data we collected about the students. Some popular machine learning classifiers are support vector machines, k-nearest neighbors, random forests, neural networks, and NB. The NB classifier is one of the most straightforward and most accurate data-classifying algorithms. The base of this classifier is the well-known Bayes theorem, used in probability theory. The simplicity, accuracy, and robustness of NB have made NB a popular classifying technique with various applications [32]. As the literature indicates, NB is one of the top-performing classifiers used in data mining [33].

4.4.2 Decision tree algorithm

A decision tree is a sequential technique used in machine learning. It is made of three parts: decision nodes, branches, and the end nodes, called leaves, representing the outcomes. The decision tree operates like a flow chart to make the decision-making process of classifying a new object to its appropriate type or class variable. This classification process starts from the top of the tree to the bottom. Decision trees have been used as a classification tool across many domains. This model has numerous applications in students' performance prediction and classifications.

A study conducted by Pandey and Sharma [34] used various decision tree models to predict student's performances. In a similar study, [35] utilized several decision tree models to predict students' academic success. The literature indicates various applications of machine learning models in educational data mining, but it is hard to see the application in elementary student's performance analysis. In this study, given the students' characteristics, we use the decision tree model to predict whether a student can meet the benchmark for the DIBELS score.

4.4.3 Implementation of machine learning models

Machine learning models are applied in a standard pattern, as illustrated in Fig. 1. As it shows, the data set is first preprocessed, where things like handling missing values and data scaling are implemented. After that, the dataset is divided into training and testing data. Usually, 70% of the data are randomly selected for the training, and the remaining 30% are used for the testing data. Next, a candidate model is proposed and trained on the training data before it is tested using the testing data. The candidate model is fine-tuned by using hyper-parameterization, and finally, the best model is selected.

Fig. 1
figure 1

Application of machine learning model

Full size image

4.4.4 Implementation of the multi-level classification

This section aims to develop a machine-learning model to classify students into their respective academic groups based on student-related features. As we do not have existing groups, students are classified into four groups based on their DIBELS scores, and the model is developed to predict the new students’ group. The student is classified into the “Well Below (WB)” group if the student’s score is lower than the first quartile. If the score is between the first and the second quartiles, the student is classified into the group "Below, (B).” Similarly, if the student’s score is between the second and the third quartile, the student is classified into the “Above, (A)” group, while those who have a score above the third quartile are classified in the “Well Above, (WA)” group. We use the NB classifier to develop this classification model.

After randomly splitting the data into training and testing, the training dataset is used to train the model. In the testing dataset, the class variable (appropriate group) is predicted using the student-related variables and compared with the actual values in the class variable. Unlike the binary classification, the class has four levels (WB, B, A, and WA). Therefore, assessing the model’s accuracy is slightly different due to the multi-class situation. Hence, sensitivity and specificity are defined, as shown in Tables 1 and 2. In addition, each group does not have an equal number of subjects, which means we have an unbalanced group here. The balanced accuracy can be used as a performance indicator in such a situation.

Table 1 Actual and Predicted Classes
Full size table
Table 2 Sensitivity and Specificity
Full size table

4.4.5 Performance indicators of the classification model

When evaluating machine learning models, the confusion matrix plays a significant role. Let us assume the class variable has two possible values (binary classification), namely the positive and negative. As Fig. 2 shows, the confusion matrix summarizes the values of the actual and the predicted class variables (positive/negative). Most performance indicators of machine learning models are defined on the confusion matrix. This study uses the following performance indicators: accuracy, balance-accuracy, sensitivity, and specificity.

$${\text{Accuracy}} = \left( {{\text{TP}} + {\text{TN}}} \right)/\left( {{\text{TP}} + {\text{FP}} + {\text{TN}} + {\text{FN}}} \right)$$
$${\text{Precision }} = {\text{ TP}}/\left( {{\text{TP}} + {\text{FP}}} \right)$$
$${\text{Sensitivity}} = {\text{TP}}/\left( {{\text{TP}} + {\text{FN}}} \right)$$
$${\text{Specificity}} = {\text{TN}}/\left( {{\text{TN}} + {\text{FP}}} \right)$$
$$Balanced\,Accuracy\, = \,\frac{{\left( {Sensitivity\, + \,Specificity} \right)}}{2}$$
Fig. 2
figure 2

Confuision Matix

Full size image

Accuracy is a metric to quantify the classification of a model, which represents the ratio between the total positive predictions and the total instances. The precision indicates the proportion of true positive values and the total positives. Recall, or the sensitivity, quantifies the ratio between predicted and actual positives. Another indicator, specificity, represents the ratio between the total predicted and the actual negatives.

In this dataset, the proportion of positives (failing students) and negatives (passing students) are not equally distributed. This situation is considered an unbalanced situation. When working with unbalanced data, the balance accuracy, the average of the sensitivity and specificity, is regarded as a better performance indicator than sensitivity and specificity. Therefore, we calculate the balance accuracy as well.

5 Results

  1. A.

    Influential Factors on Elementary Students’ Academic Performance

  2. A

    Descriptive statistics

For the first graders, the number of hours students watch TV per day is 1.42 h (SD = 0.77), the average number of hours the students read books is 1.30 h (SD = 0.62), the average number of hours students sleep per day is 8.48 h (SD = 0.55), the average number of hours students involve in physical activates is 1.82 h (SD = 0.94). In addition, the average BMI value is 17.88 (SD = 4.05) and the average DIBELS score is 160.29 (SD = 66.64). As far as the average DIBELS score is concerned, white students have maintained 159.74 (SD = 71.03). For the black and other students the values were > 154.40 (SD = 45.82) and 163.61 (SD = 61.08) respectively.

For the second graders, the number of hours students watch TV per day is 1.49 h (SD = 0.62), the average number of hours the students read books is 1.24 h (SD = 0.51), the average number of hours students sleep per day is 8.60 h (SD = 0.65), the average number of hours students involve in physical activates is 3.12 h (SD = 1.39). In addition, the average BMI value is 18.52 (SD = 3.54), and the average DIBELS score is 226.19 (SD = 79.76). Furthermore, the average DIBELS score for white, black, and other students is 227.08 (SD = 69.56), 219.11 (SD = 89.13), and 233 (SD = 97.22), respectively.

When the distribution of DIBELS score is considered, it is approximately symmetrically distributed, while BMI scores have a skewed to right distribution. These behaviors are the same for both graders. More descriptive statistics can be seen in the Table 1. According to the collected data, 54% of the students are boys, and about 56% of the first graders are boys. Among the students, 67% are white, 14% are black, and 19% are others. More descriptive statistics can be found in Tables 3 and 4.

  1. 2.

    ANOVA results

Table 3 Descriptive statistics of grade 1 students
Full size table
Table 4 Descriptive statistics of grade 2 students
Full size table

In order to identify the impact of gender and duration of sleep for the first grader and second graders, two-way ANOVA was conducted. In this calculation, we converted the duration of sleep to a categorical variable as LESS (sleeping time is less than 9 h) and GOOD (sleeping time is more than 9 h), as explained earlier. As Table 5 indicates, for the first graders the main effects of the Gender (F(1,88) = 5.160), p < 0.05) and Sleeping time (F(1,88) = 7.415), p < 0.01) on DIBELS were observed.

Table 5 ANOVA findings for grade 1
Full size table

Table 6 indicates that LSD post hoc test results revealed that the girls have significantly higher DIBELS scores compared to the boys. Furthermore, Students who sleep more than 9 h (Good) show significantly higher DIBELS scores than the other group of first grades who sleep less than 9 h.

Table 6 Tukey HSD test for grade 1
Full size table

As Table 7 indicates, for the second graders the main effects of BMI (F(2,90) = 2.68), p < 0.1) on DIBELS were observed. According to the Table 8, it indicates that LSD post hoc test results revealed that students in the Healthy group show higher DIBELS scores than the other two groups though they are not significant.

  1. 3.

    Results of the multiple regression model.

Table 7 ANOVA findings for grade 2
Full size table
Table 8 Tukey HSD test for grade 2
Full size table

For first graders, the number of hours they sleep significantly predicted their DIBELS score, β = 41.89, p < 0.01. Gender also explained a significant proportion of variance in DIBELS scores, β = 37.13, p < 0.05. Further information can be seen in Table 9.

Table 9 Impacting variables for the first graders
Full size table

Figure 3 shows the impact of gender and sleep duration on the DIBELS is illustrated. First graders who sleep more than 9 h show higher DIBELS scores than the students who sleep less than 9 h. In the meantime, girls ‘performance on the DIBELS exam is higher than boys.

Fig. 3
figure 3

Impact of Sleep and Gender on DIBELS for First Graders

Full size image

For the second graders, the number of hours they read books significantly predicted their DIBELS score, β = 29.14.89, p < 0.05. Students BMI value also explained a significant proportion of variance in DIBELS scores, β = − 4.00, p < 0.05. Table 10 indicates the findings of the regression analysis.

Table 10 Impacting variables for second graders
Full size table

Figure 4 shows the relationship among the DIBELS score, gender, and BMI class. Based on this figure, second graders whose BMI class is in a healthy state show higher performance on the DIBELS exam than the other two classes of Obese and Close to obese.

  1. 4.

    Results of the decision tree model.

Fig. 4
figure 4

Impact of BMI and Gender on DIBELS for Second Graders

Full size image

In order to PREDICT success with DIBELS, we proposed a decision tree model. This model aims to predict whether students can meet the benchmark level of the DIBELS exam or not. Two models for grade 1 and grade 2 were constructed based on the decision tree algorithm. The model for first graders was tested on the testing dataset for grade 1, and an accuracy of 60% and recall of 71% were observed. The recall value tells us how well the model accurately identifies students who pass the DIBELS out of all test takers. The final decision tree for grade 1 is represented in Fig. 5.

Fig. 5
figure 5

Decision tree for grade 1

Full size image

Similarly, the other model for the second graders was trained and tested using the second grader's data. This model showed an accuracy of 83% and a recall of 75%. The final optimal decision tree for grade 2 is represented in Fig. 6.

  1. 5.

    Results of the Naïve Bayes (NB) Model

Fig. 6
figure 6

Decision tree for grade 2

Full size image

In order to predict students’ appropriate DIBELS level (group), an NB model was constructed. This NB classifier was trained using a randomly sampled 70% of the collected dataset. The remaining 30% of the collected data was used for the testing data. According to the outcomes, the WB group has the highest sensitivity of 92%, meaning 92% of the students who belong to the WB group can be identified accurately. The specificity of the WA group is 100%. This states that 100% of the students who do not belong to the WA group can be correctly identified as not belonging to the WA group. Similarly, the specificity percentages for B, WB, and A groups are 95, 75, and 81%, respectively.

In this dataset, the number of students belonging to each group (class) is not equal. Therefore, for unbalanced classes, a better measurement of classification accuracy is balanced accuracy. According to the outcomes, WB, B, A, and WA's balanced accuracy is approximately 84, 73, 65, and 63%, respectively.

6 Discussion

This study had three main objectives. First, we wanted to investigate the impact of various elementary students’ related factors on their academic performance. Secondly, to develop a model to forecast their academic performance. Finally, we devised a mechanism to classify these students into appropriate skilled-level groups using the identified influential factors. In this study, we used the DIBELS score as an indicator for elementary students to quantify their academic performance. This score not only quantifies an individual student’s basic literacy skills but can also be used as an indicator to compare an individual’s academic level beyond the school level.

According to the regression analysis, both the number of hours a student sleeps, and their gender statistically influence their DIBELS score. While higher hours of sleep indicate higher academic success, girls show higher performance than boys as far as the DIBELS score is concerned. Further analysis using ANOVA procedures consolidates the above relationship. Furthermore, post hoc comparisons using the Tukey HSD test indicate that the difference between the mean DIBELS score for girls and boys is significantly different (M = 4.497, p < 0.026). In addition, using the Tukey HSD test shows that there is a significant statistical difference in DIBELS scores between those who sleep more than 9 h and those who sleep less than 8 h (M = 43.056, p = 0.007). The literature indicates that [36] found that sleep problems and patterns affect elementary students’ school performance. According to the report released by the World Health Organization in 2004 [37], lack of sleeping time can impede the development of the brain and the acquisition of cognitive and intellectual capabilities. Furthermore, [38] report that inadequate sleep can negatively impact students' academics and physical and psychosocial health.

According to the regression analysis, second graders, both the number of hours students read books and the BMI values significantly impact their DIBELS score. While the duration of book reading positively (β = 29.139, p < 0.05) impacts the DIBELS score, the BMI value negatively (β = − 4.002, p < 0.05) impacts the DIBELS score. Further analysis using ANOVA indicates that.

For the second graders, the main effects of the BMI (F(1,90) = 2.68), p < 0.1) on DIBELS were observed. As Kuczmarski [39] state, a child of this age group is considered “Healthy” if the BMI range is between 13.6 and 17.5 or in the 5th to 85th percentile of the BMI values. Furthermore, a child is considered “Close to obese” if the BMI values are between 17.5 and 19 (or 85th to 95th percentiles). If the BMI value is more than 19 (or over the 95th percentile), the child is considered “Obese.” According to the results of the ANOVA procedure, though there is no statistically significant difference among the three above groups, students in the Health group showed higher performance in DIBELS scores than the other two. This proves the findings of a study conducted by [40] about a negative association between BMI scores and student's academic achievements.

Furthermore, according to the World Health Organization (WHO), childhood obesity influences premature death and disability in adulthood [41]. As the Centers for Disease Control and Prevention (CDC) indicates, childhood obesity is a complex health issue that harms the body in various ways. Morrison [42] reveals that obesity has caused higher anxiety, depressive symptoms, and poorer quality of life. Furthermore, [43] and [44] state that low self-esteem, lower self-reported quality of life, and social problems, such as bullying and stigma, are other consequences of childhood obesity.

The second objective of this study was to predict student's success with the DIBELS exam. Using their related variables, we used the decision tree approach to predict whether students can meet the benchmark of the DIBELS exam or not. According to the experimental outcomes, if students’ related variables such as gender, the number of hours they spend sleeping, reading books, and watching TV, their race, and their BMI values are known, our constructed model can accurately predict students' success with the DIBELS exam.

Our final objective of this study was to forecast students’ appropriate academic groups at an early stage. We used another machine learning model, Naïve Bayes, for this classification, and the experimental findings show promising results. In this process, we group the students into four levels, WB, B, A, and AB, based on the students’ DIBELS score. According to the trained Naïve Bayes model, students’ appropriate academic groups can be identified accurately using the students’ related information.

This model can be beneficial for administrators in identifying students who need extra attention to succeed in their studies and identifying students who are currently performing well. In either case, the administrators can allocate resources effectively to improve the student’s success. Furthermore, the improvement in the placement of a student into an appropriate learning group enables teachers to use relevant pedagogy and pacing to improve student learning.

7 Conclusion

This study was conducted with three aims: to observe the relationship between students' related variables and academic performances, to develop a model to predict students meeting the level of academic success, and to group students into uniform groups based on their academic levels. Students’ academic success was quantified using the DIBELS score, which represents student's early literacy skills.

Using the collected dataset of first and second-graders student-related variables (characteristics), we devised techniques to achieve the above three aims. According to the experimental findings, first-grade students’ academic success is significantly impacted by the number of hours of sleep and their gender. This shows the association of higher academic success with a higher duration of sleeping time. In addition, the second graders’ academic performance is significantly influenced by the BMI value and the number of hours of book reading. These results confirm some of the findings of the literature as well as the recommendations given by the World Health Organization regarding the required number of hours of sleep for elementary students.

Another objective of this study was to develop a model to predict whether students can meet the benchmark level of academic success at the end of the year. Using the machine learning model based on decision trees, we constructed a model to predict the students’ end-of-year academic performance successfully. Finally, we constructed another accurate machine learning model called Naïve Bayes to classify students into similar groups based on their endemic levels. These models are significantly crucial for parents, teachers, and school administrators. Parents can pay attention to the variables associated with the kids that impact their academic performances, such as sleeping, book reading times, and BMI values. Furthermore, teachers and administrators can use the developed models to predict students’ future academic success before they reach the end of the semester. The above-constructed model can be effectively used if the administrators want to identify students of similar academic levels to form separate classes.

The authors have identified some limitations of a study like this. The data of this type of study can show some inaccuracies due to the parent-reported data such as sleeping time, TV watching time, reading time, and physical activity time. In this study, we used DIBELS values as the indicator for academic success, but if there are alternative measurements, this proposed model can also be applied to them. Unfortunately, we could not collect data from other graders from elementary school. Analyzing other elementary students’ data in a future study would be interesting.