I start the day by attending my regular classes and check in with my professors if I need assistance or did not understand something.
Fresh from her IQP (Interactive Qualifying Project) in Iceland, Alissa began seeking a research project where she could continue to use the skills [giving presentations, leading conversations, meeting people in professional settings] she had honed during her global projects experience.
In her junior year, Alissa initiated an Independent Study Project with Professor Rodica Neamtu and Vestigo Ventures, a venture capital firm that was looking to work with a WPI student. As part of her project, she developed a Python-based tool that uses machine learning to classify a website as a financial technology website or a non-financial technology website. The tool she created achieves a maximum classification accuracy of 96%, and it significantly reduces the time needed to manually classify a website.
"The first component of our system is a web crawler, a program which automatically browses pages within a company’s website using particular keywords, such as 'About,' 'Product,' and 'Motivation.' We extracted the text from each of these pages. Now, we want to represent the text in a numerical form so the machine learning algorithm can “read” it. We aimed to give each word a numerical weight which represents a word’s usefulness in describing a particular company website. To do this, we extracted TF-IDF, or term-frequency inverse document frequency, features from the text," she explains.
"This technique gives a higher weight to words that are unique to 'fintech' websites and a lower weight to words found in the majority of websites in the dataset. For each website, we have a large TF-IDF feature matrix which represents the website content numerically. After extracting features from each website, we labeled the feature matrices as fintech and non-fintech based on the website classes in our training data. The matrices, along with their labels, are inputted to a support vector machine, which creates a hyperplane to separate datapoints into fintech and non-fintech classes based on the training data," she adds.
Vestigo integrated her tool into their data processing workflow, and Alissa and her team wrote a paper about their work that is currently under review for the Institute of Electrical and Electronics Engineers (IEEE) Big Data conference.
Alissa, who will continue to work with the company for her MQP (Major Qualifying Project), also presented about her project with Vestigo research during the Arts & Sciences Week Undergraduate Student Lightning Talks session.
“Professor Neamtu has been one of the most supportive professors I've had at WPI," says Alissa."She always pushes me to succeed and improve my work. She is extremely dedicated and invested in our research projects, and I am very grateful for all the feedback and support she's given me to this day. She encourages me to step outside my comfort zone, be confident in my abilities, and explore new frontiers in research.”
After graduation, Alissa plans to pursue a PhD in machine learning. She hopes to use what she’s learned to develop AI-driven technologies that can be used in everyday life.
Rodica Neamtu, Computer Science
- Received the Two Towers Prize (2019)
- Gave a Lightning Talk on research with Vestigo during Arts & Sciences week
- Wrote a paper that was submitted to the Institute of Electrical and Electronics Engineers (IEEE)
- Music (she plays flute, piano, and guitar)
On Tuesday and Thursday afternoons, I attend my military classes and ROTC lab. On days when I do not have ROTC, I usually go to the gym, play intramural sports [volleyball], or catch up with friends in the Campus Center.
At the end of the day, I usually catch up on any academic work that I didn't finish in between classes during the day.