Classification of Fire and Smoke Images using

2022 14th International Conference on Mathematics, Actuarial Science, Computer Science and Statistics (MACS) | 978-1-6654-6071-2/22/$31.00 ©2022 IEEE | DOI: 10.1109/MACS56771.2022.10022449

Decision Tree Algorithm in Comparison with

Logistic Regression to Measure Accuracy,
Precision, Recall, F-score
*

B. Haranadh Reddy Karthikeyan P R

Department of Electronics and Communication Engineering Department of Electronics and Communication Engineering
Saveetha School of Engineering Saveetha School of Engineering
Chennai, Tamilnadu. India Chennai, Tamilnadu. India
[email protected] [email protected]

Abstract—The study’s objective is to assess how well the for fire detection. [4]. [1] explains the Decision tree-based
decision tree algorithm and the logistic regression algorithm Fire predicting system based on the weather elements such
classify photographs of fire and smoke. A total of 4,000 photos as Temperature, Humidity. [40] propose a fire prediction
are captured, of which 2,000 are of fire and the remaining
2,000 are of smoke.A training dataset (n=3200, or 80% of the using IoT sensors in a combination of machine learning
dataset) and a validation dataset (n=800, or 20% of the dataset) algorithms. [40] suggests that a boosted decision tree is better
are each created.Classification is carried out using Python’s than other algorithms with Auc is 0.78. [33] determine the
Sklearn machine learning package. Metrics like precision, recall, decision tree and other machine learning algorithms’ potential
f score, and accuracy numbers are used to assess an algorithm’s for mapping flammability and fire prediction. Best citation is
performance. In the classification of photos of fire and smoke,
the Novel Decision Tree provides 90.54 percent accuracy whereas [1].Intelligent chatbots, online recommendation engines, smart
logistic regression provides 85.75 percent accuracy (p 0.001). home systems, and other applications all use machine learning
Compared to the logistic regression algorithm, the decision tree (ML) techniques.Prior to this, our team has extensive expertise
algorithm performed noticeably better. working on numerous research projects across several fields.
Index Terms—Keywords: Novel decision tree, logistic regres- [5], [6], [9], [13], [22], [23], [26], [28], [29], [31], [34], [37]–
sion, fire and smoke detection, machine learning
[39] [3], [7], [36].We were inspired to explore this idea by the
I. INTRODUCTION expanding trend in this field. The accuracy of classification
is subpar in the research that is currently available. Typically,
Fire detection is the process of identifying fires like a forest
point sensors such as smoke sensors, temperature sensors, and
fire, indoor fire, and industrial fire accidents, etc. The fire may
humidity sensors are used by fire detectors to identify any
be caused due to natural hazards or by some human mistakes.
fire incidents.These devices have drawbacks, that they can
It is highly important to predict the fire at the earliest to
detect fire only when the smoke, temperature, and alternative
prevent loss of lives and other damage to properties [14],
physical characteristics reach their threshold value. This may
[27]. If there is less accuracy in the classification of fire and
cause huge damage and takes a long time to recover from the
smoke [17] it leads to more false alarm rate. This study helps
damage caused by it. In this paper, we studied the classification
decrease the false alarm rate by improving accuracy in the
of fire and smoke using the novel decision tree and the logistic
classification of fire and smoke images. Classification of fire
regression algorithms.
and smoke can be used in fire fighting robotic vehicles, and in
many surveillance applications for prediction of and detection II. MATERIAL AND METHOD
of fire and smoke [24], [20]. In the previous five years, a large
The Department of Electronics and Communication Engi-
number of research articles on the classification of fire and
neering at Chennai, India’s Saveetha School of Engineering is
smoke photos using decision tree algorithms were released.
where this research is being done. The photos used to construct
66 research papers are published in IEEE Explore in addition
the dataset were downloaded from Kaggle, GitHub, and a
to the 217 research articles that are published in Science Direct
number of other sources. The gathered pictures are split into
and the 7980 that are published in Google Scholar. A computer
two categories (fire and smoke images).The dataset consists
vision system based on image processing is now employed
of a total of 4000 photographs, 2000 of which are smoke
Identify applicable funding agency here. If none, delete this. images for group 2 and 2000 of which are fire images, re-

978-1-6654-6071-2/22/$31.00 ©2022 IEEE

Authorized licensed use limited to: VIT University- Chennai Campus. Downloaded on January 07,2025 at 05:54:04 UTC from IEEE Xplore. Restrictions apply.
spectively. [4].800 images make up the test dataset (calculated
with g power of 0.8 and alpha of 0.05; Kane, Phar, and BCPS,
n.d.), while the training dataset has 3200 images. The dataset
contains original colour photos that have been downsized to
224 x 224 pixels and turned into grayscale versions. The
test dataset comprises of 400 fire images and 400 smoke
images in the same ratio as the training dataset, which consists
of 1600 fire images and 1600 smoke images.A supervised
machine learning method called logistic regression uses a set
of independent factors to predict the values of categorical
dependent variables. [12]. It gives probabilistic values for any Fig. 1. The decision tree algorithm’s structure. The root node of the decision
real value in the range of 0 to 1. In logistic regression, there tree is followed by a decision node and a leaf node. Making decisions is done
is a concept of the threshold value. It classifies the classes via decision nodes, which also have many branches and leaf nodes as their
results.The blue box represents the leaf nodes, the yellow box the decision
based on a threshold value. The values below the threshold nodes, and the red box the root node.
belong to 0 and values above the threshold value tend to 1.
Logistic regression works based on the logistic function are
also called a Sigmoid function. Fit the training set to the 5). Entropy quantifies impurity using attribute values from
Logistic Regression. Logistic regression models are trained equation 4.
with training set images. Analyze the test results by predicting H(S) = −Pyeslog2Pyes − Pnolog2Pno (4)
the test images using logistic regression and obtaining the Σ Si
confusion matrix. Determine the Performance metrics using informationgain = Entropy(S)− entropy(S ) (5)
i
the Confusion matrix. Logistic regression is based on the S
following equations (1, 2 and 3). P yes represents a probability of one class, and P no
y = b0x0 + b1x1 + b2x2 + ......... + bnxn (1) represents a probability of a different class, where H is entropy.
In the CART (Classification and Regression Tree) algorithm,
1 a decision tree is built using the Gini index (equation 6), a
Sigmoidf unctionp = (2) measure of impurity or purity. It’s only appropriate for binary
1 + e−y splits.
Σk

log( GI = 1 − Pi2 (6)

p
) = b0x0 + b1x1 + b2x2 + ........... + bnxn (3)
1−p i=0
The decision tree method enables result classification and In this proposed work, google Colab open source platform has
prediction. Classification issues are the principal usage for the been used with core i5 processor and 8GB RAM. The built in
decision tree. Some noticeable types of decision tree algo- functions for decision tree and logistic regression algorithms
rithms are ID3 Iterative Dichotomiser developed by Quinlan are used from the python programming library. Eighty percent
and C4.5 is implemented from ID3 and next one is CART of the data from the total sample size were utilised for
(Classification and Regression tree) algorithm is the third type training, while the remaining twenty percent were used for
of Decision tree algorithm (Abid and Izeboudjen 2020). CART testing. Every statistical analysis is carried out using SPSS
is mainly applicable for Binary splits. In this work, the CART 23. (SPSS Inc., Chicago, Illinois, USA). Different algorithms
algorithm is applied for the classification of images. The basic are used to carry out descriptive statistics (mean, standard
structure of a decision tree is shown in Fig. 1. The input deviation, and standard error).The effectiveness of the logistic
to a training algorithm may be a set of labeled data and regression algorithm and the decision tree method have been
therefore the output may be two classes, fire, and smoke, compared using an independent-samples T test.The features
the generated tree is employed for the classification of new in the data set are the independent attributes in this case,
input. The training set is fixed for the decision tree algorithm. and the accuracy, recall, precision, and f-score values are the
The root node is selected based on the attribute values.The dependent variables.
technique’s optimal attribute values are referred to as the
Attribute Selection Measure (ASM). Use these attribute values III. RESULTS
to create subsets of the dataset. This ASM can be measured The mean accuracy value, f-score, precision and recall for
using two popular techniques: Information gain and Gini the decision tree algorithm are 90.54 ± 0.8773, 90 ± 0.9278,
index. By using the attribute values the data is divided into 90.9 ± 0.9135 and 89.70 ± 1.007 respectively and for the
subsets and decision tree nodes are generated. In the proposed logistics regression algorithm are 85.75 ± 0.7240, 84.87 ±
technique Gini index is used to measure the attribute values. 0.9278, 86.44 ± 0.4926 and 83.37 ± 0.9807 respectively (Table
This process is continued until the final node is reached. After 2). The outcomes of the algorithms can be quantified by
a dataset has been segmented based on a characteristic, the using metrics like precision, recall, F-score and accuracy. The
change in entropy is measured as information gain (equation following four parameters must be determined in order to

Authorized licensed use limited to: VIT University- Chennai Campus. Downloaded on January 07,2025 at 05:54:04 UTC from IEEE Xplore. Restrictions apply.
 TABLE I TABLE II
FOR THE DECISION TREE (DT) AND LOGISTIC REGRESSION (LR) ACCURACY, PRECISION, RECALL, AND F SCORE IN CLASSIFICATION OF
APPROACHES, THE MEAN ACCURACY VALUE, STANDARD DEVIATION, AND FIRE AND SMOKE IMAGES USING DECISION TREE AND LOGISTIC
STANDARD ERROR MEAN ARE COMPUTED USING DESCRIPTIVE REGRESSION METHODS.
STATISTICS.
Algorithm Precision Recall F score Accuracy
N mean Std.deviation Std. error Logistic regression 86.4 83.6 85 86.15
Accuracy LR Decision Tree 90.7 90 90.4 90.9
20 85.75 0.72394 0.16188
20 90.54 0.87729 0.19617
DT
TABLE III
F score LR INDEPENDENT SAMPLE TEST FOR SIGNIFICANCE AND STANDARD ERROR
20 84.87 0.6851 0.15333
DETERMINATION.
20 90.00 0.9278 0.0748
DT Leven’s Test for Equality of Variance t-test for Equality of Variance 95% Confidence Interval of the difference
Precision LR Equal Variance assumed F
2.051 sig.
0.160 t
18.854 dif
38 Sig 0.000
(2-tailed) Std.Error Difference
0.2543 lower
5.3075 upper
4.2788
20 86.44 0.49258 0.11015 Accuracy
Equal variance not assumed 18.854
16.581
36.64
38
0.000
0.000
0.2543
0.2307
5.6082
4.3178
4.2778
4.2778
20 90.9 0.91351 0.20427
Equal Variance assumed
F score 6.694 0.014
Equal variance not assumed
DT
16.581 29.18 0.000 0.2307 4.3225 4.2778
Equal Variance assumed 20.165 38 0.000 0.3142 6.9735 4.2778
Precision 0.906 0.332
Equal variance not assumed
Recall LR
20.165 37.97 0.000 0.3142 6.9765 4.2788
19.848 38 0.000 0.2579 5.6481 4.5982
20 83.37 0.98073 0.21930 Recall
Equal variance assumed
Equal variance not assumed
0.085 0.085
19.848 34.98 0.000 0.2579 5.6481 4.5982

20 89.70 1.00067 0.22511

DT
A. Statistical Analysis
is carried out using the SPSS programme. In order to
analyse this work, an independent sample test is used. a
comparison of the classification of fire and smoke photographs
using the decision tree algorithm and the logistic regression
algorithm.the statistically significant difference between the
evaluation of the logistic regression algorithm and the decision
tree algorithm. Over logistic regression, the decision tree’s
precision has a mean that is higher (91.34). (86.53). The
decision tree’s recall has a mean that is higher than that of
the logistic regression (89.7) (83.37). The decision tree has
the highest F score mean (90) than the logistic regression
F score mean (84.87). The main evaluation metric is accu-
Fig. 2. Accuracy of Decision tree for different Train and test Sizes.The
decision tree is depicted in red, and logistic regression is shown in blue. racy.The decision tree algorithm’s accuracy is (90.54), which
is higher than logistic regression’s accuracy (85.75). Table 2
provides an illustration of the descriptive statistical analysis.
estimate these metrics: true positive (TP), true negative (TN), Calculations are made on the logistic regression and decision
false positive (FP), and false negative (FN). The mathematical tree algorithms’ classification performance means, standard
equations listed below are used to evaluate performance (7, 8, deviations, and differences that are statistically significant.
9 and 10) p 0.001 indicates a substantial difference between the two
TP groups. The independent sample test complete analysis is listed
Precision = (7) in Table 3. Bar chart comparing the mean (+-1SD) precision,
TP + FP recall, f-score, and accuracy of logistic regression algorithm
and decision tree algorithm. There is a significant difference
Recall = TP (8) between the two groups p¡0.001. From the results obtained it
TP + FN is clear that the accuracy and other performance metrics of the
decision tree are higher than the logistic regression which is
F − score =(2precision∗recall) (9) shown in Fig. 3. Independent sample test was used to compare
(precision+recall) the accuracy of two algorithms and a statistically significant
difference was noticed P¡0.001.
Accuracy = T P +T N (10)
(T P +T N +F P +F N )
IV. DISCUSSION
Table 1 displays the results of the logistic regression and Analysis is done on the effectiveness of the logistic re-
decision tree algorithms in terms of precision, recall, f-score, gression and decision tree algorithms. Both decision trees
and accuracy. For various test and train sizes, Fig. 2 displays and logistic regression have been evaluated for precision,
the decision tree and logistic regression’s accuracy. When recall, f-score, and accuracy. Following result analysis, it
trained with more photos, the innovative decision tree performs was shown that the decision tree approach outperformed the
more accurately. Logistic regression is not as accurate as the logistic regression. From the data, it is evident that the decision
decision tree technique. tree has superior accuracy, f score, precision, and recall

Authorized licensed use limited to: VIT University- Chennai Campus. Downloaded on January 07,2025 at 05:54:04 UTC from IEEE Xplore. Restrictions apply.
 V. CONCLUSION
Accurate and earlier detection of forest fire and smoke can
reduce the loss of lives and preserves the natural resources
in the forest. When it came to identifying photographs of fire
and smoke, the decision tree method outperformed the logistic
regression technique by a wide margin.
REFERENCES
[1] Faroudja Abid and Nouma Izeboudjen. Predicting Forest Fire in
Algeria Using Data Mining Techniques: Case Study of the Decision
Tree Algorithm. Advances in Intelligent Systems and Computing, 2020.
[2] Abdulwahab Ali Almazroi, Hitham Alamin, Radhakrishnan Sujatha, and
Noor Zaman Jhanjhi. A web-based model to predict a neurological
disorder using ann. In Healthcare, volume 10, page 1474. MDPI, 2022.
[3] Saif EA Alnawayseh, Waleed T Al-Sit, Taher M Ghazal, et al. Smart
congestion control in 5g/6g networks using hybrid deep learning tech-
niques. Complexity, 2022, 2022.
[4] Nahid Chowdhury, Dewan Ruhul Mushfiq, and Azm Ehtesham Chowd-
Fig. 3. Mean comparison of classification performances in terms of accuracy, hury. Computer vision and smoke sensor based fire detection system. In
f score, precision, and recall of logistic regression and decision tree. 2019 1st International Conference on Advances in Science. Engineering
and Robotics Technology (ICASERT, 2019.
[5] Anil Kumar Danda, M. R. Muthusekhar, Vinod Narayanan, Mirza F.
Baig, and Avinash Siddareddi. Open versus closed treatment of unilateral
than the logistic regression depicted in Fig. 3.The decision subcondylar and condylar neck fractures: A prospective, randomized
tree approach performs better for classification than logistic clinical study. Journal of Oral and Maxillofacial Surgery: Official
Journal of the American Association of Oral and Maxillofacial Surgeons,
regression. [18] 80% accuracy in wildfire prediction is shown 68(6):1238–41, 2010.
using five decision tree-based classifiers to analyse the spatial [6] Sherwin I. DeSouza, M. R. Rashmi, Agalya P. Vasanthi, Suchitha Maria
distribution of wildfires. With a 67.3% accuracy, [41] were Joseph, and Rashmi Rodrigues. Mobile phones: The next step towards
healthcare delivery in rural india? PloS One, 9:8, 2014.
able to predict the occurrence of fires using data from forest [7] Abebe Diro, Haftu Reda, Naveen Chilamkurti, Abdun Mahmood, Noor
fires and a classification model developed using the C4.5 Zaman, and Yunyoung Nam. Lightweight authenticated-encryption
decision tree technique. [30] conducted a study and used scheme for internet of things based on publish-subscribe communication.
IEEE Access, 8:60539–60551, 2020.
the spatial entropy-based decision tree technique to create a [8] Varun Dogra, Sahil Verma, Kavita Verma, NZ Jhanjhi, Uttam Ghosh, and
decision tree to forecast hotspot occurrences in the Bengkalis Dac-Nhuong Le. A comparative analysis of machine learning models
region of the Riau province of Indonesia. The results of this for banking news extraction by multiclass classification with imbalanced
datasets of financial news: Challenges and solutions. 2022.
work using the 5-fold cross-validation test are decision trees [9] Devaraj Ezhilarasan, Velluru S. Apoorva, and Nandhigam Ashok Vard-
with an average accuracy of 89.04% on the training set and han. Syzygium cumini extract induced reactive oxygen species-mediated
52.05% on the testing set. [10] explained the prediction apoptosis in human oral squamous carcinoma cells. Journal of Oral
Pathology Medicine: Official Publication of the International Associa-
of ignition of fire and transmit the data through a wireless tion of Oral Pathologists and the American Academy of Oral Pathology,
sensor network. Decision Tree analyzes four meteorological 48(2):115–21, 2019.
parameters: temperature, relative humidity, wind speed, and [10] Demin Gao, Jie Xin, and Fuquan Zhang. A Decision Tree Algorithm for
Forest Fire Prediction Based on Wireless Sensor Networks. International
daily precipitation, while the other part is the prediction results Journal of Embedded Systems, 2020.
of the Forest Weather Index system. The Proposed Decision [11] Sukanta Ghosh, Amar Singh, Kavita, NZ Jhanjhi, Mehedi Masud, and
Tree algorithm is easy to understand because it follows the Sultan Aljahdali. Svm and knn based cnn architectures for plant classi-
fication. CMC-COMPUTERS MATERIALS & CONTINUA, 71(3):4257–
human thinking ability while making decisions and it is easier 4274, 2022.
to understand than the logistic regression because it shows a [12] Y. Jafari Goldarag, Y. Jafari Goldarag, Ali Mohammadzadeh, and A. S.
tree-like structure. Logistic regression is mainly applicable for Ardakani. Fire Risk Assessment Using Neural Network and Logistic
Regression. Journal of the Indian Society of Remote Sensing, 2016.
linear data. Nonlinear data is not very good. Our institution [13] Lokhasudhan Govindaraju, Prasanna Neelakantan, and James L. Gut-
is passionate about high quality evidence based research and mann. Effect of root canal irrigating solutions on the compressive
has excelled in various fields [?], [2], [8], [9], [11], [15], strength of tricalcium silicate cements. Clinical Oral Investigations,
21(2):567–71, 2017.
[16], [19], [21], [25], [32], [34], [35], [42]–[45]. We hope this [14] Nizar Hamadeh, Ali Karouni, and Bassam Daya. Predicting Forest
study adds to this rich legacy. The limitation of this research Fire Hazards Using Data Mining Techniques: Decision Tree and Neural
work is that the decision tree may undergo overfitting issues Networks. Advanced Materials Research, 2014.
[15] Mamoona Humayun, Noor Z Jhanjhi, Abdullah Almotilag, and
and computational complexity while doing classification. The Maram Fahhad Almufareh. Agent-based medical health monitoring
decision tree was very complex because it contains a lot of system. Sensors, 22(8):2820, 2022.
layers. Utilize five decision tree-based classifiers to analyse [16] Mamoona Humayun, R Sujatha, Saleh Naif Almuayqil, and NZ Jhanjhi.
A transfer learning approach with a convolutional neural network for the
the spatial distribution of wildfires and show that wildfire classification of lung carcinoma. In Healthcare, volume 10, page 1058.
prediction is 80% accurate. With a 67.3 percent accuracy The MDPI, 2022.
obtained accuracy can be improved by using other types of [17] Iswanto Iswanto, Kunnu Purwanto, Weni Hastuti, Anis Prabowo, and
Muhamad Yusvin. Smart Smoking Area Based on Fuzzy Decision Tree
decision trees such as ID3, C5, and boosted decision tree Algorithm. International Journal of Advanced Computer Science and
algorithms. Applications, 2019.

Authorized licensed use limited to: VIT University- Chennai Campus. Downloaded on January 07,2025 at 05:54:04 UTC from IEEE Xplore. Restrictions apply.
[18] Abolfazl Jaafari, Eric K. Zenner, and Binh Thai Pham. Wildfire Spatial [39] C. H. Sekhar, V. Narayanan, and M. F. Baig. Role of antimicrobials
Pattern Analysis in the Zagros Mountains, Iran: A Comparative Study in third molar surgery: Prospective, double blind,randomized, placebo-
of Decision Tree Based Classifiers. Ecological Informatics, 2018. controlled clinical study. The British Journal of Oral Maxillofacial
[19] Upasna Jindal, Surjeet Dalal, G Rajesh, Najm Us Sama, NZ Jhanjhi, Surgery, 39(2):134–37, 2001.
and Mamoona Humayun. An integrated approach on verification of [40] Richa Sharma, Shalli Rani, and Imran Memon. A Smart Approach for
signatures using multiple classifiers (svm and decision tree): A multi- Fire Prediction under Uncertain Conditions Using Machine Learning.
classification approach. 2021. Multimedia Tools and Applications, 2020.
[20] J. Kamalakannan, Anurag Chakrabortty, Gaurav Bothra, Prateek Pare, [41] Imas Sukaesih Sitanggang and Mohd Hasmadi Ismail. Classification
and C. S. Pavan Kumar. Forest fire prediction to prevent environmental Model for Hotspot Occurrences Using a Decision Tree Method. Geo-
hazards using data mining approach. In Proceedings of the 2nd matics, Natural Hazards and Risk, 2011.
International Conference on Data Engineering and Communication [42] Gokul Sridharan, Pratibha Ramani, Sangeeta Patankar, and Rajagopalan
Technology, 2019. Vijayaraghavan. Evaluation of salivary metabolomics in oral leukoplakia
[21] MO Khairandish, M Sharma, V Jain, JM Chatterjee, and NZ Jhanjhi. and oral squamous cell carcinoma. Journal of Oral Pathology Medicine:
A hybrid cnn-svm threshold segmentation approach for tumor detection Official Publication of the International Association of Oral Pathologists
and classification of mri brain images. IRBM, 2021. and the American Academy of Oral Pathology, 48(4):299–306, 2019.
[22] Muthusamy Senthil Kumar, Gelli Vamsi, Ramasamy Sripriya, and [43] Kathiravan Srinivasan, Lalit Garg, Debajit Datta, Abdulellah A Al-
Praveen Kumar Sehgal. Expression of matrix metalloproteinases (mmp- aboudi, NZ Jhanjhi, Rishav Agarwal, and Anmol George Thomas.
8 and -9) in chronic periodontitis patients with and without diabetes Performance comparison of deep cnn models for detecting driver’s
mellitus. Journal of Periodontology, 77(11):1803–8, 2006. distraction. CMC-Computers, Materials & Continua, 68(3):4109–4124,
[23] Seungjin Lee, Azween Abdullah, Nz Jhanjhi, and Sh Kok. Classification 2021.
of botnet attacks in iot smart factory using honeypot combined with [44] Imran Taj and Noor Zaman. Towards industrial revolution 5.0 and
machine learning. PeerJ Computer Science, 7:e350, 2021. explainable artificial intelligence: Challenges and opportunities. Interna-
[24] Mohammad Sultan Mahmud, Md Shohidul Islam, and Md Ashiqur tional Journal of Computing and Digital Systems, 12(1):295–320, 2022.
Rahman. Smart Fire Detection System with Early Notifications Using [45] Sobia Wassan, Chen Xi, Nz Jhanjhi, and Hassan Raza. A smart
Machine Learning. International Journal of Computational Intelligence comparative analysis for secure electronic websites. INTELLIGENT
and Applications, 2017. AUTOMATION AND SOFT COMPUTING, 30(1):187–199, 2021.
[25] M. G. Mathew, S. R. Samuel, A. J. Soni, and K. B. Roopa. Evaluation
of Adhesion of Streptococcus Mutans, Plaque Accumulation on Zirconia
and Stainless Steel Crowns, and Surrounding Gingival Inflammation in
Primary Clinical Oral Investigations, 2020.
[26] Meenu Mehta, Devesh Tewari Deeksha, Gaurav Gupta, Rajendra
Awasthi, Harjeet Singh, Parijat Pandey, et al. 2019:206–15, August.
[27] Mikhail D. Molovtsev and Irina S. Sineva. Classification algorithms
analysis in the forest fire detection problem. 2019, 2019.
[28] Iffat Nasim, Prasanna Neelakantan, R. Sujeer, and C. V. Subbarao. Color
stability of microfilled, microhybrid and nanocomposite resins–an in
vitro study. Journal of Dentistry, 38:137–42, June 2010.
[29] P. Neelakantan, C. Subbarao, C. V. Subbarao, G. De-Deus, and M. Zehn-
der. The impact of root dentine conditioning on sealing ability and
push-out bond strength of an epoxy resin root canal sealer. International
Endodontic Journal, 44(6):491–98, 2011.
[30] Indry Dessy Nurpratami and Imas Sukaesih Sitanggang. Classification
Rules for Hotspot Occurrences Using Spatial Entropy-Based Decision
Tree Algorithm. Procedia Environmental Sciences, 2015.
[31] Saurav Panda, Jayakumar Doraiswamy, Sankari Malaiappan, Sheeja Saji
Varghese, and Massimo Del Fabbro. Additive effect of autologous
platelet concentrates in treatment of intrabony defects: A systematic re-
view and meta-analysis. Journal of Investigative and Clinical Dentistry,
7(1):13–26, 2016.
[32] J. Pc, T. Marimuthu, and P. Devadoss. Prevalence and Measurement of
Anterior Loop of the Mandibular Canal Using CBCT: A Cross Sectional
Study. Clinical Implant Dentistry and Related Research, 2018.
[33] Binh Thai Pham, Abolfazl Jaafari, Mohammadtaghi Avand, Nadhir Al-
Ansari, and Tran Dinh Du. Hoang Phan Hai Yen. Tran Van Phong, et al
“Performance Evaluation of Machine Learning Methods for Forest Fire
Modeling and Prediction.” Symmetry, 2020.
[34] Neeraja Ramadurai, Deepa Gurunathan, A. Victor Samuel, Emg Subra-
manian, and Steven J. L. Rodrigues. Effectiveness of 2. 2019.
[35] Asha Ramesh, Sheeja Varghese, Nadathur D. Jayakumar, and Sankari
Malaiappan. Comparative estimation of sulfiredoxin levels between
chronic periodontitis and healthy patients - a case-control study. Journal
of Periodontology, 89(10):1241–48, 2018.
[36] Abdul Razaque, Ahmed Haroon, Manoj Kumar, Gullapalli Amulya,
and Tariq Rahim Soomro. Smart bandwidth friendly buffer: Handling
overflow in wireless mesh networks. Journal of Independent Studies
and Research, 14(1):22, 2016.
[37] D. Sajan, K. Udaya Lakshmi, Y. Erdogdu, and I. Hubert Joe. Molecular
structure and vibrational spectra of 2,6-bis(benzylidene)cyclohexanone:
A density functional theoretical study. Spectrochimica Acta. Part A,
Molecular and Biomolecular Spectroscopy, 78(1):113–21, 2011.
[38] Arumugam Sathivel. Hanumantha rao balaji raghavendran. Periasamy
Srinivasan, and Thiruvengadam Devaki “Anti-Peroxidative and Anti-
Hyperlipidemic Nature of Ulva Lactuca Crude Polysaccharide on D-
Galactosamine Induced Hepatitis in Rats, 46(10):3262–67, 2008.

Authorized licensed use limited to: VIT University- Chennai Campus. Downloaded on January 07,2025 at 05:54:04 UTC from IEEE Xplore. Restrictions apply.