Abdullah Albyati Portfolio

Medical Center Network Design

Abdullah Albyati

Design Wired and Wireless Network for a Medical Center with 8 Buildings

This project provided a medical center an operational network design which will digitize the medical record system and upgrade the centers current LAN network to a secure wireless network that covers the entire medical center campus area.

Checkout the Project Here

2018 WORLD CUP QUARTERFINALS PREDICTION

The Dashboard

In the PowerBI dashboard above I analyzed the data below to represent the probability of each team winning their next game based on their performance in the world cup so far. Read below to learn how I created this prediction.

The Data

I collected the data from Fox Sports Stats Page After careful consideration of which features to use I decided on the features in the table below to build my machine learning model (Some of my past soccer experience played a role on deciding which data is more important)

team	Goals Scored	Goals Conceded	Wins	Attempts	Attemps on Target	Pass Accuracy	Distance Covered (MI)	W/L
Russia	9	5	3	33	12	0.78	296	W
France	7	4	3	45	17	0.84	253	W
Croatia	8	2	3	61	16	0.83	274	W
Brazil	7	1	3	77	29	0.88	256	W
Sweden	6	2	3	48	18	0.77	260	W
Belgium	12	4	4	77	30	0.87	261	W
England	9	4	3	58	17	0.89	283	W
Uruguay	7	1	4	50	18	0.85	260	W
Colombia	6	3	2	39	12	0.81	271	L
Denmark	3	2	1	40	12	0.81	288	L
Argentina	6	9	1	53	18	0.86	248	L
Japan	6	7	1	42	16	0.85	260	L
Mexico	3	6	2	57	13	0.85	245	L
Portugal	6	6	1	52	14	0.84	257	L
Spain	7	6	1	70	21	0.91	279	L
Switzerland	5	5	1	56	14	0.85	265	L

The Code

Below is the Python code to connect to the Microsoft Azure Machine Learning Experiment follow the instruction in the code for more info.

The Algorithm

In my prior prediction model that I used to predict the round-of-16 match winners I used decision tree regression model to calculate each team chance of winning. My model predicted 7 winners out of 8 right!

For the quarterfinals I experimented with few algorithms in Microsoft Azure Machine Learning Studio and I decided on using Two-Class Neural Network model to predict the chance of winning for each team. Classification using neural networks is a supervised learning method, and therefore requires a tagged dataset, which includes a label column. For example, you could use this neural network model to predict binary outcomes such as whether or not a patient has a certain disease, or whether a machine is likely to fail within a specified window of time.

" A neural network is a set of interconnected layers. The inputs are the first layer, and are connected to an output layer by an acyclic graph comprised of weighted edges and nodes. Between the input and output layers you can insert multiple hidden layers. Most predictive tasks can be accomplished easily with only one or a few hidden layers. However, recent research has shown that deep neural networks (DNN) with many layers can be very effective in complex tasks such as image or speech recognition. The successive layers are used to model increasing levels of semantic depth. The relationship between inputs and outputs is learned from training the neural network on the input data. The direction of the graph proceeds from the inputs through the hidden layer and to the output layer. All nodes in a layer are connected by the weighted edges to nodes in the next layer. To compute the output of the network for a particular input, a value is calculated at each node in the hidden layers and in the output layer. The value is set by calculating the weighted sum of the values of the nodes from the previous layer. An activation function is then applied to that weighted sum."

In order to make sure the algorithm works I added a Win/Loss column to the data set and used that column as the labeled data column and all other columns (except team name) as the features data. When running the model in Microsoft Azure the algorithm predict all 8 teams as winners and add a scored label column (probability of winning) to the data set as an output. Check out the full Microsoft Azure Machine Learning experiment

The Prediction

The algorithm takes into account all columns (features) except the team name of the data set to train and predict the probability of winning (W/L column) after running the experiment all you need to do is feed it the required numbers and it will give each team a percentage (probability of winning). As you can see in the PowerBi dashboard at the beginning of the page all teams (except Russia) has about 90% chances of winning each game. That's why I left the result in decimal points to know which team has a slightly higher chance of winning. Choose the 2 teams that are playing against each other and the team with a higher probability should be the winner (According to the gods of Data!)

TWITTER TEXT ANALYSIS WITH PYTHON

Project Summary

Background
Aliens exist! Scientists have intercepted a communications stream from another planet for the first time. Initial analysis has shown that the data stream is in fact the aliens’ version of Twitter.
The United Nations have decided that we will reach out to these aliens and contact them, but only after we have learned as much as we can about them. At present we cannot translate their language, but we can try to learn about them as a society and as individuals by textual analysis of the stream.
The data
The data to analyze is saved in two text files:

follows.txt which contains the information about which other users follow each user; and
stream.txt which is the actual alien Twitter stream of tweets.

2.1 Usernames
In the text files a user name is a sequence of alphanumeric characters, length >= 0; for example,
mrmagoo23.

2.2 follows.txt
The file follows.txt contains the follow graph. Each line is of the form

user0<>user1<>user2<>...

where <> represents some whitespace. It means that user0 follows user1, user2, ... If a user doesn’t follow anyone, the line looks like this:

user0

The lines are in no particular order.

2.3 stream.txt
The file stream.txt contains the Twitter stream in chronological order. Each line is one of:

ordinary tweet

user0<>any text...

where user0 is the username of the author of the tweet, and any text... is any sequence of characters terminated by the end of the line, the content of the tweet.
retweet

user0<>RT<>@user1<>any text...

where user0 is the username of the user who has retweeted an original tweet authored by user1.
direct message

user0<>DM<>@user1<>any text...

where user0 is author of a private message, normally seen only by user1.

Within the content of any tweet (any text...) can appear mentions of other users. Mentions start with
@ followed by a user name. The user name is terminated by the end of the line or any non-alphanumeric character. Any user mentioned will see that tweet, even if it is a direct message. A tweeter might make a mistake, so a mentioned user, or direct message addressee, might not actually exist.

Tasks and Solutions
Your tasks for this assignment are to write Python programs that analyze one or more of the data files to answer some questions we have about the alien Twitter users.New Paragraph

3.1 Task 1: Who are the most popular users?
The more popular a user is, the more other users follow the user. Write a program that prints the user name of the most popular user. In the case of a tie, print the user names of all the most popular users, one per line, in lexicographic order. Sample output:

andrew mrmagoo23 sally

My solution:

3.2 Task 2: Who are the top parrots?
Write a program that prints the usernames of the top n parrots and how many tweets they retweeted. n is a number prompted for and entered by the user. The top parrots retweeted the most. Sample output for n = 5:

1034 mrwordy
999 chatty3 d555lucy
10 blabberMcBlabberFace john mrmagoo23

Note:

• how the output is formatted in a table with justified columns;
• that users with the same numbers of tweets were printed in lexicographic order; and
• that even though only the top 5 were requested, because of a tie, more than 5 were printed.

My solution:

3.3 Task 3: Who are the worst trolls?
Write a program that prints the usernames of the worst n trolls and how many times they mentioned other users. n is a number prompted for and entered by the user. The worst trolls are those who mentioned the most other users in their tweets, including addresses in direct messages and the authors of retweeted tweets. Format the output in the same manner as for task 2.

My solution:

3.4 Task 4: Who are the biggest influencers?
Write a program that prints the usernames of the top n users than on average have the highest number of users that see each of their tweets, and the average number of times each tweet was seen. A user will see a tweet if it comes from a user they follow, or they are mentioned in the tweet, including as addresses in direct messages and as the authors of retweeted tweets. Sample output (note the formatting):
1034.6 popidol34 vacuousTwit
999.0 smartymcsmartyface
10.4 chatty3 d555lucy zippo

My solution:

CROATIA vs FRANCE PREDICTION

The Dashboard

In the PowerBI dashboard above I analyzed the data below to represent the probability of each team winning their next game based on their performance in the world cup so far. Read below to learn how I created this prediction.

The Data

I collected the data from https://www.foxsports.com/soccer/fifa-world-cup/stats
The prediction model was trained using data from all teams performances in the competition so far. and by supplying the 2 finalist stats below the model predicted the probability of winning based on their overall performance in the world cup.

I also used the same model to predict the winning team based on their performance in the semi-final match and in both instances, the winner was France (I really want Croatia to win though). Below is the prediction based on semi-final stats

Team	Goals Scored	Goals Conceded	Wins	Attempts	Attempts on Target	Pass Accuracy	Distance Covered (MI)	Scored Probabilities
France	10	4	4	75	24	0.84	377	0.999131441
Croatia	10	4	3	78	19	0.83	361	0.990501881

team	Goals Scored	Goals Conceded	Attempts	Attempts on Target	Pass Accuracy	Distance Covered (MI)	Scored Probabilities
France	1	0	19	5	0.86	63	0.979853511
Croatia	2	1	22	7	0.79	89	0.966043413

The Code

Below is the Python code to connect to the Microsoft Azure Machine Learning Experiment follow the instruction in the code for more info.

The Algorithm

In the prior prediction model that I used to predict the round-of-16 match winners I used decision tree regression model to calculate each team chance of winning. My model predicted 7 winners out of 8 right!

For the quarterfinals I experimented with few algorithms in Microsoft Azure Machine Learning Studio and I decided on using Two-Class Neural Network model to predict the chance of winning for each team. Classification using neural networks is a supervised learning method, and therefore requires a tagged dataset, which includes a label column. For example, you could use this neural network model to predict binary outcomes such as whether or not a patient has a certain disease, or whether a machine is likely to fail within a specified window of time.

" A neural network is a set of interconnected layers. The inputs are the first layer and are connected to an output layer by an acyclic graph comprised of weighted edges and nodes.

Between the input and output layers you can insert multiple hidden layers. Most predictive tasks can be accomplished easily with only one or a few hidden layers. However, recent research has shown that deep neural networks (DNN) with many layers can be very effective in complex tasks such as image or speech recognition. The successive layers are used to model increasing levels of semantic depth.

The relationship between inputs and outputs is learned from training the neural network on the input data. The direction of the graph proceeds from the inputs through the hidden layer and to the output layer. All nodes in a layer are connected by the weighted edges to nodes in the next layer.

To compute the output of the network for a particular input, a value is calculated at each node in the hidden layers and in the output layer. The value is set by calculating the weighted sum of the values of the nodes from the previous layer. An activation function is then applied to that weighted sum."

In order to make sure the algorithm works I added a Win/Loss column to the data set and used that column as the labeled data column and all other columns (except team name) as the features data. When running the model in Microsoft Azure the algorithm predicts all 8 teams as winners and add a scored label column (probability of winning) to the data set as an output. Check out the full Microsoft Azure Machine Learning experiment.

WORLD CUP 2018 PREDICTION

The Dashboard

In the dashboard above I analyzed the data from the teams' performance during the group stage of the world-cup to predict the winner of each game during the round of 16.

The Data

I collected the data from https://www.foxsports.com/soccer/fifa-world-cup/stats
After careful consideration of which features to use I decided on the features in the table below to build my machine learning model (Some of my past soccer experience played a role on deciding which data is more important)

I also utilized data from https://data.world/sawya/football-world-cup-2018-dataset using the SQL query extract the players from the qualified teams.

teams	GF	GA	W	attempts	attemps on target	passes accuracy	distance covered
Russia	8	4	2	27	11	0.8	206
Spain	6	5	1	45	12	0.91	194
Portugal	5	4	1	32	9	0.85	191
France	3	1	2	45	17	0.84	195
Denmark	2	1	1	25	9	0.82	204
Croatia	7	1	3	39	9	0.84	192
Argentina	3	5	1	45	10	0.8	190
Brazil	5	1	2	56	19	0.89	195
Switzerland	5	4	1	38	10	0.86	201
Sweden	5	2	2	36	15	0.78	195
Mexico	3	4	2	44	12	0.85	185
Belgium	9	2	3	53	22	0.87	194
England	8	3	2	42	15	0.91	194
Colombia	5	2	2	25	8	0.82	187
Japan	4	4	1	31	12	0.86	192
Uruguay	5	0	3	44	15	0.87	194

The Code

Below is the Python code to connect to the Microsoft Azure Machine Learning Experiment follow the instruction in the code for more info.

The Algorithm

I used Microsoft Azure Machine Learning Studio to create the model. It was a trial and error experiment to decide which regression model will give the right prediction values. I decided to use a decision forest regression model from Microsoft Azure Machine Learning Studio algorithms.

Decision trees are non-parametric models that perform a sequence of simple tests for each instance, traversing a binary tree data structure until a leaf node (decision) is reached.

Decision trees have these advantages:
They are efficient in both computation and memory usage during training and prediction.
They can represent non-linear decision boundaries.
They perform integrated feature selection and classification and are resilient in the presence of noisy features.
This regression model consists of an ensemble of decision trees. Each tree in a regression decision forest outputs a Gaussian distribution as a prediction. An aggregation is performed over the ensemble of trees to find a Gaussian distribution closest to the combined distribution for all trees in the model.

The image below shows the simple model from Microsoft Azure Machine Learning Studio and here is the link to the experiment on Azure AI Gallery https://gallery.cortanaintelligence.com/Experiment/World-Cup-2018-Round-of-16-prediction

The Prediction

The algorithm takes into account all columns (features) except the team name of the dataset to train and predict the chances of winning (W column) after running the experiment all you need to do is feed it the required numbers and it will give each team a percentage (probability of winning).

Let’s take for example Spain vs Russia match, Spain was favorite to win but the machine learning algorithm predicted Russia with 1.7 chance of winning and Spain with 1.4 chance of winning. Russia actually won the game as my model predicted.

The same scenario worked with Argentina vs France where many favorite Argentina to win but my model predicted France to win. My prediction model takes into account the team performance in the group stage and uses those features to predict the outcome.

HURRICANE IRMA SHELTERS PROJECT

Abdullah Albyati

Provide Interactive Map To Help People Find Shelters During Hurricane Irma

Get the Geo-location (longitude and latitude) of each shelter by using their physical address. The code below uses R ggmap package and will work on any project just copy the code and run it on your environment. When running the code it will ask for a file to read, make sure to provide a file with a column named addresses.

Checkout the App Here

The Data

During hurricane Irma, I was stationed in Hawaii but I wanted to help the preparation effort as much as I can, and since the hurricane path was so hard to predict, Florida's emergency services opened many shelters scattered around Florida.

After hours of research, I realized there was no one list that includes all the shelters and there was no easy way to find the shelters other than going to Florida's emergency response agency website which provided lists per county and each county has a different website.

Long story short, I decided to compile all locations on one list (you can download it here) and create an interactive map that includes all shelters and when the users click on a shelter they will automatically be redirected to Google maps website or app to navigate to the shelter. Download the data here

The Code

There were two major steps implemented to finish coding this map.
Get the Geo-location (longitude and latitude) of each shelter by using their physical address. The code below uses R ggmap package and will work on any project just copy the code and run it on your environment. When running the code it will ask for a file to read, make sure to provide a file with a column named addresses.

The second step of this project is to take the data set created from the first step and utilize plotly, ggmap, and leaflet packages to create the map and link the address to their Google maps location.

SWIFTKEY WORD PREDICTION PROJECT

Abdullah Albyati

Use text Data TO predict the Next Word

This is the Capstone project for Johns Hopkins University’s Data Science specialization on coursera.org The goal of this capstone project is to build a Shiny application that is capable of predicting the next word based on user text input.

Checkout the App Here

The Interface

I used shinydashboard R package in RStudio to create the layout for the app. In addition, I used shinydashboardthemes package to create the theme, Shiny Dashboard Themes package is still in beta but it worked perfectly for the purpose of this app.

The Data

The dataset for this project contains over 4 million lines of text divided into 3 text documents (news, blogs, and Twitter) it was provided as part of the final capstone project for Johns Hopkins University's Data Science Specialization on coursera.org in partnership with Swiftkey. The data can be downloaded here

The Code

There were three steps taken to code the project with R. First step was analyzing the large corpus of text documents to discover the structure in the data and how words are put together to better understand the overall data. See below Exploratory Analysis section which includes all the code for the first step.

The Prediction Model

AUTOMATIC DA-2062 GENERATOR

Abdullah Albyati

making my job and others easier

To my Army friends, I made this app a while back to help me print out 2062s for all Soldiers in my company. Now I have hosted it in the cloud, and anyone can use it. If you want to issue out equipment to Soldiers like masks, M4s or whatever and you have a list of names and assigned SNs you can copy and paste, and this web app will make a DA-Form 2062 with the info you entered for every Soldier.

Checkout the App Here

Summary

During my time in the Army I had to issue out equipment to Soldiers during training events the process was tedious, to be honest; I will have 300+ Soldiers that need to fill out a DA-Form 2062 Hand Receipt and I have to input the equipment serial number that the Solider is assigned on the form for each individual.

So, I decided I need to automate this process to make it faster and more efficient. I started this project by making an excel macro (see image below)

You can download the excel file and use it if you rather using the excel version of this website. Download the DA-FORM2062 Excel with Macro below is the Visual Basic code that I used to automate the excel file.

Unfortunately, governemnt computers don't allow Excel Macros, so I decided to make the web app you see at the top of the page. I didn't know how to code with JavaScript before this app so I had to learn JavaScript for this project.

I used Nodejs as the built environment for this JavaScript web application. Below is the app.js code

Hello, I am

Abdullah Albyati

I never guess. It is a capital mistake to theorize before one has data. Insensibly one begins to twist facts to suit theories, instead of theories to suit facts.

My Education Journey

Project Highlight

Experience

Security Analyst | Google

Special Projects Analyst | Palo Alto Networks

Data Analyst | Uber

Data Analyst | Sensploration Coffee

Information Management Officer | US Army

Team Leader | U.S Army

Program Assistant | Catholic Charities

Linguist | GLOBAL LINGUIST SOLUTIONS

Senior Network Engineer | Al-Naba Co

Portfolio

Skills

100 %

50 %

75 %

90 %

90 %

80 %

80 %

60 %

60 %

50 %

90 %

Certifications

Oracle Database SQL Certified Associate

CIW Site Development Associate

CompTIA Linux+ Certified

Data Science - Johns Hopkins University

Data Scientist with R - Data Camp

CCNA Routing and Switching - Cisco

CCNA Security - Cisco

CNSS 4011 Recognition - Cisco

My Journey

CONTACT ME