Analysis & Detection of autism spectrum disorder using ML (3)_RKS.pptx
1. ANALYSIS & DETECTION OF
AUTISM SPECTRUM
DISORDER USING ML
Presented By :
Roshan Kumar(19BTRCS061)
Prathamesh Samdadiya(19BTRCS065)
Santhosh Raj R(19BTRCS069)
Guided By : Dr A. Prakash
2. • Autism Spectrum Disorder (ASD) is a neurological disorder which might
have a lifelong impact on the language learning, speech, cognitive, and
social skills of an individual.
• Its symptoms usually show up in the developmental stages, and it
impacts around 3% of the population globally .
• The main objective is to thus determine if the child is susceptible to ASD
in its nascent stages using ML models, which would help streamline the
diagnosis process.
2
ABSTRACT :
3. 1. Does the child make eye contact with others?
2. Does the child respond to their name when called?
3. Does the child point or use gestures to communicate?
4. Does the child have difficulty engaging in reciprocal social
interactions with others?
5. Does the child show an interest in playing with toys or objects
in an unusual manner?
3
QUESTION:-
4. 6. Does the child have difficulty understanding or using
language appropriately for their age?
7. Does the child have difficulty following social cues or
understanding social rules?
8. Does the child engage in repetitive behaviors or have
restricted interests?
9. Does the child have difficulty adapting to changes in
routine or transitioning between activities?
10. Does the child exhibit any unusual sensory behaviors,
such as avoiding certain textures or sounds?
4
QUESTION:-
5. LITERATURE REVIEW :
5
Several studies have made use of machine learning in various ways to improve and speed up the
diagnosis of ASD. Duda et al applied forward feature selection coupled with under sampling to
differentiate between autism and ADHD with the help of a Social Responsiveness Scale containing 65
items. Deshpande et al. used metrics based on brain activity to predict ASD. Soft computing
techniques such as probabilistic reasoning, artificial neural networks (ANN), and classifier
combination have also been used . Many of the studies performed have talked of automated ML
models which only depend on characteristics as input features. A few studies relied on data from
brain neuroimaging as well. This was used to classify between patients with and without ASD,
respectively. Thabtah et al. proposed a new ML technique called Rules-Machine Learning (RML) that
offers users a knowledge base of rules for understanding the underlying reasons behind the
classification, in addition to detecting ASD traits. Al Banna et al. made use of a personalized AI-
based system which assists with the monitoring and support of ASD patients, helping them cope with
the COVID-19 pandemic.
In this study, we have used five ML models to classify individual subjects as having ASD or No-ASD,
by making use of various features, such as age, sex, ethnicity, etc., and evaluated each classifier to
determine the best performing model.
7. COMPARISONS BEFORE AND AFTER COVID-19
* According to a Survey
7
• Paired t-tests revealed decreases
in both outdoor recreation
participation (64% reported
declines) and subjective well-being
(52% reported declines).
• These Social Impacts have led to
growth of Autism in several parts
of the world
8. WORKING MODEL -
• Figure demonstrates the general working and flow
of our system. We begin by preprocessing the
dataset to eliminate missing values and outliers,
remove noise, and encode categorical attributes.
• This reduces data dimensionality to improve speed
and efficiency during training.
• Once the data set has been preprocessed,
classification algorithms like Logistic Regression,
Naïve Bayes, Support Vector Machine,
Convolutional Neural Network, and Random Forest
Classifiers are used to predict the output label (ASD
or no ASD).
9. 9
METHODOLOGY
• Data Pre-processing - Preprocessing refers to the
transformations applied to a data set before feeding it to the
model. It is done to clean raw or noisy data and make it
more suited for training and analysis.
• Classification Algorithms - After having
performed data preprocessing , we applied five
classification models, namely
❑ Logistic Regression
❑Naive Bayes
❑Support Vector Machine
❑Convolutional Neural Network
❑Random Forest Classifier
10. ANALYSIS AND RESULTS
10
• We plotted several graphs to get different
visual perspectives of the dataset.
• In the first plot we can see that the number
of toddlers who are ASD positive is those
who do not have jaundice while birth.
• The count is over 2 times that of jaundice
born toddlers.
• Thus, we can infer that jaundice born
children have a weak link with ASD.
11. 11
• For toddlers, most of the ASD
positive cases happen to be at
are around 36 months of age.
• The least number of cases
were observed between 15
and 20 months of age. From
the graph, it is evident that
significant signs of autism
occur at the age of 3 years .
• According to data , one out of
every 68 children aged
between 2 and 3 years has
autism.
12. NECESSITY OF THIS PROJECT -
12
• The assessment of ASD behavioral traits is a time taking process that is
only aggravated by overlapping symptomatology.
• There is currently no diagnostic test that can quickly and accurately detect
ASD, or an optimized and thorough screening tool that is explicitly
developed to identify the onset of ASD.
• We have designed an automated ASD prediction model with minimum
behavior sets selected from the diagnosis datasets of each.
• Out of the five models that we applied to our dataset; Logistic Regression
was observed to give the highest accuracy.
• The primary limitation of this research is the scarce availability of large
and open source ASD datasets. To build an accurate model, a large dataset
is necessary. The dataset we used here did not have sufficient number of
instances.
13. REFERENCES
13
1. Dataset:https://www.kaggle.com/fabdelja/autism-screening-for-
toddlers.
Accessed 1 Oct 2019.
2. Thabtah, F. An accessible and efficient autism screening method for
behavioral data and predictive analyses. Health Informatics Journal
2019;25(4):1739–55. https://doi.org/10.1177/1460458218796636.
3. Vaishali, R., and R. Sasikala. "A machine learning based approach to
classify Autism with optimum behavior sets. (2018) " International
Journal of Engineering & Technology 7(4).
4. Fadi Thabtah. (2017). “Autism spectrum disorder screening: machine
learning adaptation and DSM-5 fulfillment.” In Proceedings of the 1st
International Conference on Medical and Health Informatics.
5. Sarfaraz Masood, Adhyan Srivastava, Harish Chandra Thuwal, and
Musheer Ahmad. (2018). “Real-time sign language gesture (word)
recognition from video sequences using CNN and RNN.” In Intelligent
Engineering Informatics (pp. 623-632). Springer, Singapore.