IRSML Neal Kurande Wina Godwin Anyanwu Jr Adam

Team Members ● Wina. Godwin Anyanwu Jr. - 3 rd Year Computer Science Major

What is IRS? The Intelligent Response System aims to create a user interface that

Last Semester - The IRS project was split into two teams The IRS-ML team

Semester Goals 1. To analyze data using machine learning techniques 2. To modularize the

General Improvements ● Code Modularized ○ ○ File structure changed Code separated into methods

K-Means Clustering ● A common form of clustering that creates n-number of clusters from

Agglomerative Clustering ● A form of hierarchical clustering that uses a bottom-up approach ●

Challenges ● ● ● Downloading and installing Ubuntu Version Control Insufficient Data Accessing more

Next Steps ● ● Incorporating other data streams into the clustering algorithms Using a

Slides: 16

Download presentation

IRS-ML Neal Kurande, Wina. Godwin Anyanwu Jr. , Adam Chau

Team Members ● Wina. Godwin Anyanwu Jr. - 3 rd Year Computer Science Major - Experience: Java, C, Android, Python ● Adam Chau - 2 nd Year Computer Science Major - Experience: Java, Python, Java. Script, SQL ● Neal Kurande - 3 rd year Computer Engineer - Experience: C/C++, Python, Java. Script, MATLAB

What is IRS? The Intelligent Response System aims to create a user interface that leverages the ITS database to provide feedback that TA’s can use to improve student performance in target subject areas. ● This is accomplished by accessing the SQL database in python, pulling specific data, and then converting it to a json file that’s read by a GUI.

ITS Student Feedback Loop

Last Semester - The IRS project was split into two teams The IRS-ML team worked on pulling data from the database - - Accessed question score, rating, and duration The IRS-GUI team worked on creating a GUI that could visualize this data This was completed by generating json files on the backend that the front end team would then convert using REACT.

Last Semester

Semester Goals 1. To analyze data using machine learning techniques 2. To modularize the code to improve future developer experience 3. To make the system dynamic and update in realtime based on the SQL database

General Improvements ● Code Modularized ○ ○ File structure changed Code separated into methods ● Implementation Improvements ○ ○ ○ Can choose data by semester Can select data by pre and post-test (Chapters 1 - 7 & Chapter 8 respectively) Json names are generated based on the data parameters ● Used the Github Wiki ○ Wiki now exists

K-Means Clustering ● A common form of clustering that creates n-number of clusters from a dataset ● K-Means is an iterative algorithm-creates n number of clusters, finds the centroid then remakes the clusters ● K-Means needs data preprocessing ○ ○ ○ Need to first eliminate outliers from the dataset Normalize all dimensions of the dataset or create appropriate weights for each dimension Eliminate Na. N data points

K-Means Pictures

Agglomerative Clustering ● A form of hierarchical clustering that uses a bottom-up approach ● Clusters are grouped together using the euclidean distance ● Data for spring 2018 and summer 2018 used to make seven clusters for the graphs

Agglomerative Clustering

Challenges ● ● ● Downloading and installing Ubuntu Version Control Insufficient Data Accessing more data streams Choosing relevant ML algorithms

Next Steps ● ● Incorporating other data streams into the clustering algorithms Using a different type of clustering or unsupervised learning Integrating with the cloud to run clustering in real time Displaying data via the IRS-GUI

DEMO!