ANALYSIS OF COMMON SUPERVISED LEARNING ALGORITHMS THROUGH APPLICATION

Advanced Computational Intelligence: An International Journal, Vol.10, No.1/2/3, July 2023
DOI: 10.5121/acii.2023.10303 29
ANALYSIS OF COMMON SUPERVISED LEARNING
ALGORITHMS THROUGH APPLICATION
Palak Narula
Adobe Inc., Noida, India
ABSTRACT
Supervised learning is a branch of machine learning wherein the machine is equipped with labelled data
which it uses to create sophisticated models that can predict the labels of related unlabelled data.the
literature on the field offers a wide spectrum of algorithms and applications.however, there is limited
research available to compare the algorithms making it difficult for beginners to choose the most efficient
algorithm and tune it for their application.
This research aims to analyse the performance of common supervised learning algorithms when applied to
sample datasets along with the effect of hyper-parameter tuning.for the research, each algorithm is applied
to the datasets and the validation curves (for the hyper-parameters) and learning curves are analysed to
understand the sensitivity and performance of the algorithms.the research can guide new researchers
aiming to apply supervised learning algorithm to better understand, compare and select the appropriate
algorithm for their application. Additionally, they can also tune the hyper-parameters for improved
efficiency and create ensemble of algorithms for enhancing accuracy.
KEYWORDS
Decision Trees, Neural Networks, Boosting, Support Vector Machines, K-nearest neighbours
1. INTRODUCTION
ML and AI technologies are fuelling the growth of industries across the globe. There has been a
rapid transformation and focus towards using these technologies to improve the existing systems
as well as develop new systems that can understand the existing trends and predict new trends. A
vast majority of these applications require supervised learning algorithms as a part of their
complex systems. Though there exists a wide variety of algorithms that can be used for different
applications, it is difficult for a beginner to decide the best-suited algorithm for the application. If
the algorithm is not well-suited for the application, it can lead to issues (like reduced accuracy,
difficulty in maintenance, increased downtime due to learning) when the application is deployed
for actual clients.
The objective of this research is to bridge the gap between theory and real-world application of
supervised learning algorithm for the new researchers. The analysis of the algorithms is done on
two physical world applications – company bankruptcy prediction and breast cancer
classification. The common supervised learning algorithms are applied to these datasets (split into
training and testing data) and an analysed based on the validation curve for different values of
hyper-parameter, learning curve for different data sizes and finally the accuracy and RMSE
metrics of the entire data (before and after hyper-parameter tuning) [6,7]

30
2. RELATED WORK
There are some researches available that study common supervised learning algorithms and aim
to compare the algorithms. Iqbal Muhammad and Zhu Yan [1] surveyed various supervised
learning algorithms through a theoretical approach highlighting various issues with supervised
learning algorithms, common metrics for comparison and the internal working of the algorithms.
R. Saravanan and Pothula Sujatha [2] reviewed, compared and classified supervised machine
learning algorithms to give readers an overview of these algorithms in terms of data
classification. The research classifies the algorithms as linear and probabilistic classifiers and
highlights how a label is chosen for each category. Burkart, N. and Huber, M.F [3] provided essential
definitions, principles and methodologies of supervised learning algorithms and conducted a survey to
review the approaches.
The related researches have analysed the supervised learning algorithms theoretically and did not
discuss about the practical approaches to compare the algorithms and understand and improve
metrics through hyper-parameter tuning.
3. CLASSIFICATION PROBLEMS
3.1. Company Bankruptcy Prediction
As mentioned on Kaggle [4], this data was collected from Taiwan Economic journal from 1999 to
2009. This data s based on the business regulation of Taiwan Stock exchange. This data can be
used for a binary classification problem where the historical combination of these features can be
used to predict the probability of a company getting bankrupt in the future. The data includes
around 95 features (almost all features being continuous) to define the current standing of the
company in the Taiwan stock market. The data also includes two types of labels for these
companies - bankrupt and not bankrupt.
3.2. Breast Cancer Classification
This dataset is also derived from Kaggle [5]. The data represents an important binary
classification problem where the historical data can be used to understand the role of various
features in defining the diagnosis of breast cancer tissue. It includes the historical data of around
thirty features (almost all being continuous) to define whether the diagnosis of the breast tissue
should be malignant (cancerous) or benign (non-cancerous).
4. DECISION TREES
Decision trees [8] are a type of supervised machine learning algorithm where the data is split
continuously according to a certain parameters and finally the tree leaves represent the
output/label of the data. Following hyper-parameters are chosen for tuning and analysis-
• Max depth - This parameter defines the max depth of the decision tree. This parameter
forces the algorithm to prune the tree to avoid overfitting.
• Min_sample_split - This parameter defines the minimum number of samples that should
be present at a node for splitting the node.

31
4.1. Validation Curves for Breast Cancer Data
Figure 1. Validation curve for max_depth parameter
The above validation curve represents the accuracy of the decision tree classifier with different
values of max depth of the tree. The graph shows that the accuracy of the training score keeps
increasing which is expected since the classifier would try to create the maximum fitted decision
tree from the training data with each leaf representing the label.
The next plot represents the accuracy of the training data and the cross-validation data for
different values of minimum sample split. As expected, the training data has the maximum
accuracy when there is no restriction on minimum number of samples for splitting the data. The
accuracy decreases with the increased restriction.
Figure 2. Validation curve for min_sample_split parameter
4.2. Comparison
Table 1. Comparison based on hyper-parameter tuning
Metric Default Parameters Post hyper-parameter tuning
Accuracy 91.61% 93.71%
Root mean squared error 8.39% 6.29%
The Decision tree classifier with the default parameters has also been able to provide a decent
accuracy level but with the tuned hyper-parameters, (pruned decision tree and more restrictive
rules for splitting the data) there was a limited scope of overfitting the training data and hence
better accuracy.

32
4.3. Learning Curve for Breast Cancer Data
The plot represents the learning curve of Decision Tree with Breast Cancer Data. The model
seems to have a decent learning with the accuracy of cross validation curve increasing with the
increase in training data size. There is not much learning of the model post 200 sample size
which can be pointed as the optimal size for the learning of Decision tree for breast cancer data.
Figure 3. Learning curve for Breast cancer Data
4.4. Validation Curves for Bankruptcy Data
The above plot represents the validation curve for max_depth parameter of Decision tree for
Bankruptcy data. Similar to the previous dataset, the curve for training score keeps improving
accuracy with the increased value of max_depth as it tries to overfit the data.
The accuracy of the cross-validation remains constantly high upto max_depth=3 and then starts
decreasing which indicates that the Decision tree classifier can work optimally with max_depth
= 3 (which is also the value suggested by the Grid Search [13])
The validation curve in fig.5 has features similar to the validation curve for Breast cancer data.
The accuracy of the learning curve decreases with increasing restrictions in sample size for
splitting. The cross-validation curve has a consistent accuracy for different splitting values with
slightly higher value achieved at least restriction.

33
Figure 5. Validation curve for min_sample_split parameter
4.5. Comparison
Accuracy 95.48% 96.36%
The performance metrics reflects an improvement in the accuracy of the results delivered by the
Decision tree post tuning of the two hyper-parameters - max_depth and min_sample_split.
4.6. Learning Curve for Bankruptcy Data
Figure 6. Learning curve for Bankruptcy Data
The learning curve for the data shows an ever-increasing line for the cross-validation data which
indicates a good learning for the Decision tree classifier. The training data works best when the
amount of the data is low, and the classifier can overfit the data easily. The accuracy of the
training score keeps decreasing with the increasing data size. The two curves intersect at around
2250 data set size which should be the ideal data size for training the Decision tree for
Bankruptcy data.

34
5. BOOSTING
Boosting [9] is a machine learning algorithm which works by fitting a weak classifier (Decision
Tree by default) on the training data and then using the same classifier again with a higher focus
on previously incorrectly classified data. The final algorithm is a combination of these weak
classifier with adjusted weights. Following hyper-parameters are chosen for tuning-
• n_estimators - This parameter defines the maximum number of estimators where the
boosting can stop.
• learning_rate - This defines the learning rate to be applied to each classifier during the
iteration
The following plot represents the validation curve for n_estimators parameter for Breast cancer
dataset. The plot shows that the accuracy of training curve constantly increases until it reaches
the maximum accuracy and then it remains constant. This is because the boosting algorithm tries
to focus on incorrectly labelled data and with the increased number of estimators, the model tries
to overfit the data. The cross-validation curve shows an increased accuracy with the increasing
number of estimators, but the accuracy eventually decreases as the model tries to overfit the data.
There is also an increasing variance with the increasing number of estimators. So, the plot
suggests that the classifier should perform optimally for n_estimators=90.
Figure 7. Validation curve for n_estimators parameter
The following plot represents the Validation curve for learning rate parameter with Breast Cancer
dataset. The lines for both the training score and the validation scores seem to overlap for most of
the parts in the plot and have a similar variance. The plot indicates that the classifier performs
best for this data with learning_rate = 1.0
Figure 8. Validation curve for learning_rate parameter

35
5.2. Comparison
Accuracy 95.80% 96.50%
Comparing the performance metrics of Breast Cancer data with Boosting algorithm, it can be
observed that even though the algorithm uses multiple weak classifiers (Decision tree in this
case), it has a better overall performance than the decision tree. Also, the performance improves
with tuning of hyper-parameters.
The learning curve shows a continuous improvement in the learning of the model for both the
learning and the cross-validation curves. Thus, for the Breast Cancer Data, the Boosting
algorithm will perform better with the use of entire learning data.
The next plot represents the validation curve for n_estimators parameter with Bankruptcy dataset.
Similar to the plot for Breast cancer, the curve for the training data keeps improving until the
maximum accuracy is achieved indicating that the classifier is trying to overfit data with
increased number of weak classifiers.
The cross_validation curve indicates that the optimal value for this parameter for the given
dataset should be around 125.

36
Figure 10. Validation curve for n_estimators parameter
The validation curve for learning_rate for Bankruptcy dataset is similar in its properties to Breast
Cancer dataset. The classifier performs most optimal with learning_rate = 1.0 and as we
increase the weight of weak classifiers, the performance of the Boosting algorithm decreases.
Figure 11. Validation curve for learning_rate parameter
5.5. Comparison
Accuracy 96.71% 96.83%

37
The learning curve suggests that the boosting algorithm will have the most optimal learning for
data size = 2250 where the cross-validation curve peaks. For the lower data size, the algorithm
tries to overfit the data as is evident from the training score curve and for the higher data sizes,
the algorithm seems to be generalising, so the accuracy is decreasing.
6. SUPPORT VECTOR MACHINES
Support vector machines [10] are a type of machine learning algorithm that works by selecting a
hyperplane in an N-dimensional space to classify the data points. Following hyper-parameters are
chosen for tuning and analysis here-
• kernel- This parameter specifies the kernel type to be used in the algorithm. We will be
focusing on the following four types - linear, poly, rbf and sigmoid
• max_iter - This parameter defines the maximum number of iterations that the solver can
use
The next plot represents the accuracy of the SVM classifier with different available kernels for
Breast Cancer data. The curve represents four distinct values for different types of kernels so
nothing can be deduced from the shape of the curve.
For both the training and validation scores, the linear Kernel seems to be performing most
optimally followed by rbf and poly kernels and finally sigmoid doesn’t seem to fit well for the
breast cancer data. So, it seems for the Breast Cancer data, linear kernel for SVM would be the
best choice.
Figure 13. Validation curve for kernel parameter
Next, we analyse the algorithm for max_iter parameter. For the analysis, I chose the linear kernel
and plotted the validation curve for different values of max_iter parameter. The curve shows an
increasing accuracy with an increased limit in maximum number of allowed iterations. The curve
eventually flattens because the classifier starts to converge before utilising the maximum
available limit. Since the classifier is converging pretty fast, selecting max_iter = 60 safely
allows the classifier to work at its optimal capacity.

38
Figure 14. Validation curve for max_iter parameter
6.2. Comparison
Accuracy 95.10% 97.20%
The performance metrics indicates a significant improvement in the accuracy of the SVC with
parameter hyper-tuning. The accuracy is also better than the previous two machine learning
classifiers- decision tree and boosting.
6.3.Learning Curve for Breast Cancer Data
The learning curve indicates a continuous improvement in the accuracy of the classifier with the
increasing training data size. The training score is highest when the available data is low since
overfitting is possible but with increased data, the bias is reduced, and cross-validation score
keeps improving. Overall, it seems that the maximum available data should be used to train an
SVC for Breast Cancer data.

39
Figure 16. Validation curve for kernel parameter
The above curve indicates the discrete accuracy values for different possible kernels for SVC. For
the Bankruptcy data, linear kernel seems to be performing the worst. The other three kernels have
a comparable accuracy with a slightly higher values for poly and rbf.
Since the SVC with rbf kernel was converging so quickly that the graph was a straight line with
maximum accuracy from max_iter=1. So, to analyse the parameter better, I chose the worst
performing kernel, that is linear kernel and plotted the above graph for various values of
max_iter. The above validation curve shows high variance in accuracy metric for Bothe the
training and cross validation scores. There is also a great fluctuation in the accuracy of the
results. Since the linear kernel doesn’t fit well for the Bankruptcy dataset, it is important to put a
hard stop on number of iterations to avoid the classifier to run forever. In this scenario, the
fluctuations in the accuracy values seems to stabilise after max_iter = 50, so that should be value
for tuning the max_iter hyper-parameter

40
6.5. Comparison
Accuracy 96.65% 96.65%
Since the default SVC kernel is rbf and the classifier converges fast enough that the max_iter
won’t affect it’s processing so the performance metrics of default parameters is same as the
classifier with tuned hyper-parameters.
The learning curve of SVC with Bankruptcy data is quite different from the learning curves of
previous classifier. In this the cross-validation score remains constantly higher than the learning
score. This is probably because the classifier is able to fit data and converge quickly even with
the partial data, but it learns better with increased data size. So, for this dataset, SVC should
perform best with the maximum possible data size (more than 4000 data entries in this case).
7. NEURAL NETWORKS
Neural network [11] is a supervised learning algorithm inspired by the human neural system
where the input is processed by a number of neutrons arranged in layers and a relationship
between the inputs and labels is generated. Following hyper-parameters are chosen for tuning and
analysis here-
• hidden_layer_sizes - It takes tuple as an input where the ith element represents the
number of neutrons in the ith layer
• max_iter - This parameter represents the upper limit on the number of iterations that the
solver can use
The below validation curve is plotted by varying the number of neurons for a single hidden layer
in the neural network. The curve shows continuous fluctuations and high variance in the accuracy
of the classifier for lower hidden layer sizes. Eventually both the learning and crossvalidation

41
scores stabilises around hidden_layer_size=60 with the maximum score at hidden_layer_size=80
which should be the value of the hyper-parameter tuning.
Figure 19. Validation curve for hidden_layer_sizes parameter
The validation curve for max_iter parameter shows a continuous improvement in the accuracy of
the classifier with the increased value of max_iter parameter since the classifier will get a greater
number of iterations and thus will be able to perform better. Both the training and crossvalidation
curves begin to flatten around 140 so that should be the optimal value of max_iter parameter =
140
7.2. Comparison
Accuracy 93% 93%
Since the default value of both the parameters (hidden_layer_sizes and max_iter) are much higher
than our chosen values, the neural network classifier with its default hyper-parameters perform
same as the classifier with the optimal setting of the hyper-parameters. But, by setting the hyper-
parameters, a lot of unnecessary computation can be saved, and the classifier will take lesser time
to complete.

42
The learning curve shows a continuous improvement in the accuracy of the classifier with the
increase in data size but the variance decreases making the results more reliable. The curve
suggests using data size = around 260 for most optimal results with the Breast Cancer data while
using Neural Network classifier.
The validation curve for the hidden layer size parameter shows a decrease in the accuracy with
increase in the number of neurons in the hidden layer. The accuracy eventually increases to a
maximum value with some fluctuations and then starts to decrease. This behaviour of the curve is
common for both the training and cross-validation scores. The variance in the accuracy is also
high for low hidden layer size which is expected since the classifier will get fewer number of
neurons to map the input to labels correctly. Looking at the graph, the most optimal value for
hidden_layer_size parameter is 80.
Figure 22. Validation curve for hidden_layer_sizes parameter
The validation curve for max_iter parameter shows an increase in the accuracy with the increase
in the value of the parameter till max_iter=60 and then the accuracy starts decreasing and the
variance starts increasing for both the learning and cross-validation curves. The decrease in
accuracy indicates that the classifier tries to overfit data when it is allowed more restrictions so
choosing max_iter=60 should provide an optimal accuracy to the neural network classifier.

43
7.5. Comparison
Accuracy 93.19% 91.73%
The performance metrics indicate a lower performance of the classifier post tuning of
hyperparameters for bankruptcy data. This is probably because the hyper-parameter values
chosen in the process to make it optimal make the classifier much more restrictive allowing lesser
number of neurons and forcing the algorithm to converge quickly. To improve this, more
experiments can be performed with a different range for the hyper-parameter values.
The learning curve of the neural network with Bankruptcy data shows a continuous improvement
in the accuracy with an increased training size. There is also a reduction in the variance which
indicates that for complicated dataset like Bankruptcy data, a higher data size would result in
better learning for Neural Networks since it will enable the classifier to map the input to labels
better.

44
8. K-NEAREST NEIGHBOURS
K nearest neighbours [12] is a supervised learning algorithm that works by assigning label based
on class membership. The algorithm works on the belief that a data unit should have the same
label as the majority adapt units around it have.
Following hyper-parameters are chosen for tuning and analysis here-
• n_neighbours - this parameter determines the number of neighbours that should be
considered by the classifier
• metric - this parameter defines the distance function that should be used by the classifier
The below validation curve shows highest accuracy of the training curve for n = 0 and eventual
decrease in accuracy which reflects overfitting the data. The classifier tries to map each data
point to the corresponding label when it doesn’t have to check the value of its’ neighbours and
the accuracy deceases as the number of neighbours to be consulted increases. The crossvalidation
curve provides a better picture on choosing the most optimal values of n. The curve has
approximately similar from 5 to 13 with decreased values in the beginning and end of the curve.
The most optimal value from the plot is at n_neighbours = 7 which has the highest accuracy
achieved by the cross-validation curve.
Figure 25. Validation curve for n_neighbours parameter
Figure 26. Validation curve for metric parameter

45
The validation curve for metric parameter shows distinct values of the classifier for different
metrics chosen to represent the distance of the data point from its peers. Both the learning and
cross-validation curves have peak accuracy with Manhattan distance so metric=Manhattan
seems to be the most optimal setting.
8.2. Comparison
Accuracy 94.4% 95.1%
As expected, the performance metrics indicate an improvement in the accuracy of the K nearest
neighbour classifier post tuning of hyper-parameters.
The learning curve of the KNN classifier with Breast cancer data shows continuous improvement
in the accuracy of the predictions with higher slop in the beginning and a lower slope for higher
values of the data size. Since there is no continuous decrease in the performance of the classifier,
using the entire available data would be suitable for getting optimal results with KNN.
The curve below shows how the number of neighbours selected for KNN affects the accuracy of
the algorithm. As expected, the learning curve has the highest accuracy for n=0 as it tries to map
each data to the label thus overfitting the data. As the n increases, the accuracy starts decreasing.
The cross-validation curve provides a better insight about the performance of the classifier. The
accuracy of the classifier increases with increase in number of neighbours upto a value and then
stays constant. This point indicates the most optimal value of the parameter, that is,
n_neighbours=7

46
The validation curve below represents the distinct accuracy levels with different metric parameter
that can be used to measure distance of a data point from its peer. Since, the maximum accuracy
is achieved for Manhattan metric, it should be considered for tuning the hyper-parameter. But the
GridSearchCV algorithm suggest metric=Euclidean. The euclidean distance also provides a
comparable accuracy, it might be performing better when combined with other hyper-parameters
Figure 5. Validation curve for n_jobs parameter
8.5. Comparison
Accuracy 96.65% 96.65%
There is no improvement in the accuracy of the classifier by tuning of hyper-parameters probably
because we tune parameters looking at the cross-validation scores but when it comes to testing
data, the default parameters may perform as good as the tuned hyper-parameters

47
The training score in the curve shows a continuous improvement in the learning of the algorithm
with increasing data size though the cross-validation curve shows a a stable accuracy irrespective
of the data size. Honouring both the curves, choosing maximum data size for training the
classifier seems to be providing the best possible accuracy.
9. CONCLUSIONS
9.1. Summarised Analysis
• The smaller datasets are faster to train but can lead to overfitting of data. An optimised
size of dataset can be achieved by applying dimensionality reduction techniques.
• Hyper-parameter tuning helps in improving the accuracy of the algorithms, but it is a
complicated process that requires understanding the individual and combined effect of
hyper-parameter on the algorithm.
• The performance metrics should be defined before choosing an algorithm as a single
algorithm can’t satisfy all the metrics.
• An ensemble of algorithms might have better metrics than individual algorithms.
Table 11. Overall comparison of the five algorithms
Metric Best Algorithm Worst Algorithm
Accuracy (Bankruptcy data) Boosting Neural Networks
Accuracy (Breast Cancer data) SVM Neural Networks
Clock Time (Training) Neural Networks SVM
Clock Time (Testing) KNN KNN
9.2. Limitations
The research is limited to covering only a few common supervised learning algorithms with the
tuning of only two hyper-parameters per algorithm. There are a wide spectrum of application,
algorithms, hyper-parameters, and metrics that are beyond the scope of this research but are
important for more advanced analysis and understanding of the algorithms.

48
9.3. Future Scope
As future work, the analysis in this research can be used to make an informed decision while
selecting an algorithm (or an ensemble of algorithms) for a particular application.Additionally,
the research can also be taken forward to analyse new algorithms and parameters that are not
covered in this paper. Applying the research on more datasets would also help the reader
understand the nature of application and the best-suited algorithm.
REFERENCES
[1] Iqbal, Muhammad & Yan, Zhu. (2015). SUPERVISED MACHINE LEARNING APPROACHES:
A SURVEY. International Journal of Soft Computing. 5. 946-952. 10.21917/ijsc.2015.0133
[2] R. Saravanan and P. Sujatha, "A State of Art Techniques on Machine Learning Algorithms: A
Perspective of Supervised Learning Approaches in Data Classification," 2018 Second International
Conference on Intelligent Computing and Control Systems (ICICCS), Madurai, India, 2018, pp.
945-949, doi: 10.1109/ICCONS.2018.8663155.
[3] Burkart, N. and Huber, M.F. A survey on the explainability of supervised machine learning. Journal
of Artificial Intelligence Research.
[4] Fedesoriano, “Company bankruptcy prediction,” Kaggle, 13-Feb-2021. [Online]. Available:
https://www.kaggle.com/fedesoriano/company-bankruptcy-prediction.
[5] kaggle.com. (n.d.). Breast Cancer Dataset Classification. [online] Available at:
https://www.kaggle.com/dhainjeamita/breast-cancer-dataset-classification/data
[6] Mitchell, T.M. (1997). Machine learning. New York: Mcgraw-Hill
[7] omscs.gatech.edu. (n.d.). CS 7641: Machine Learning Course Videos | OMSCS | Georgia Institute
of Technology | Atlanta, GA. [online] Available at: https://omscs.gatech.edu/cs-7641-machine-
learningcourse-videos
[8] Scikit-learn.org. (2009). sklearn.tree.DecisionTreeClassifier — scikit-learn 0.21.3 documentation.
[online] Available at:
https://scikitlearn.org/stable/modules/generated/sklearn.tree.DecisionTreeClassifier.html#sklearn.tre
e.DecisionTreeCl assifier.
[9] scikit-learn.org. (n.d.). sklearn.ensemble.AdaBoostClassifier — scikit-learn 0.22.1 documentation.
https://scikitlearn.org/stable/modules/generated/sklearn.ensemble.AdaBoostClassifier.html.
[10] Scikit-learn.org. (2019). sklearn.svm.SVC — scikit-learn 0.22 documentation. [online] Available at:
https://scikit-learn.org/stable/modules/generated/sklearn.svm.SVC.html#sklearn.svm.SVC.
[11] scikit-learn.org. (n.d.). sklearn.neural_network.MLPClassifier — scikit-learn 0.24.1 documentation.
https://scikitlearn.org/stable/modules/generated/sklearn.neural_network.MLPClassifier.html#sklear
n.neural_network. MLPClassifier.
[12] scikit-learn developers (2019). sklearn.neighbors.KNeighborsClassifier — scikit-learn 0.22.1
documentation. [online] Scikit-learn.org. Available at:
https://scikitlearn.org/stable/modules/generated/sklearn.neighbors.KNeighborsClassifier.html.
[13] Scikit-learn.org. (2019). sklearn.model_selection.GridSearchCV — scikit-learn 0.22 documentation.
https://scikitlearn.org/stable/modules/generated/sklearn.model_selection.GridSearchCV.html.
AUTHOR
Palak Narula gained her BTech in Computer Science from H.B.T.I, India and MS in
Computer Science (major in Machine Learning) from Georgia Institute of Technology,
Atlanta, GA. Currently she is working as a Computer Scientist in Adobe Noida, India.

ANALYSIS OF COMMON SUPERVISED LEARNING ALGORITHMS THROUGH APPLICATION

More Related Content

Similar to ANALYSIS OF COMMON SUPERVISED LEARNING ALGORITHMS THROUGH APPLICATION (20)

Recently uploaded (20)

ANALYSIS OF COMMON SUPERVISED LEARNING ALGORITHMS THROUGH APPLICATION