Parul Saini's repositories
Social-Influence-Maximisation
The project aims to predict and maximize social media influence. The model based approach involves training the neural network using simulations to predict active node set from a given seed set. The simulations are produced using propagation models as discussed in further sections. Further LSTM based sequence to sequence models that are based on the encoder decoder architecture can be used to predict the sequence depicting the spread and extracting useful information about the network structure which is unknown. The aim of the project is to analyse social networks whose structure is not explicitly known. In case of most of the real life networks like twitter network we don't know the structure of the graph. For example we know that a tweet has been re-tweeted but we don't know the order in which the re-tweets occurred as we don't have any information regarding the spread. Being able to analyse the next active node not only provides us some useful information about the graph, but it also allows us to predict the future influence and impact of the spread.
Detecting-Depression-in-Tweets-using-Machine-Learning-Classifiers
Machine learning, when used with social media, can be useful in diagnosis of mental health illnesses as it provides insights into an individual’s behaviour. In this study, we compared several machine learning classifiers on their ability to detect if a given tweet that talks about depression actually shows signs of depression. We fetched tweets from Twitter using a hashtag keyword matching procedure and used a pre-trained BERT sentiment model to separate potential tweets about depression into two classes. Since prior research indicated that including emojis can improve classification performance when working with social media textual data, we used Word2Vec and Emoji2Vec to create embeddings for both text and emojis in the tweets. Our best performing model was a Gaussian kernel support vector machine (SVM) with a test accuracy of around 85% both with and without emojis. Contrary to our expectations, including emojis did not noticeably improve performance which we attribute primarily to our limited dataset.
Optimzing-Web-Forms-using-A-B-Testing-and-Reinforcement-Learning
Project done as a part of course CSC2558 - Topics in Multidisciplinary HCI: Designing Intelligent Self-Improving Systems Through Human Computation, Randomized A/B Experiments and Statistical Machine Learning Fall 2021 by Joseph Jay Williams
Weak-Supervision-Framework-with-Active-Learning
Machine learning models require huge amounts of labeled data in order to achieve good performance on text classification tasks. However, good quality labels are often difficult to obtain, which limits the applicability of these techniques. Weak supervision is one such data labeling technique which uses weak supervision sources for assigning noisy labels to unlabeled instances. On the contrary, active learning is a method which makes use of user input to generate good quality labels for the data. We aim to investigate how a hybrid model of the two would perform, in addition to emphasizing on the performance advantage of stand-alone active learners when compared to weak supervision frameworks. We develop an active learning module to query the most difficult instances from the unlabeled dataset and find that logistic regression classifier with uncertainty sampling is the best performing estimator - query strategy combination for active learner. We compare the performance of this active learner with various other weak supervision baseline models and identify that it outperforms them by only labeling a maximum of 100 actively sampled instances.
ADA
Some code for analysis and design of algorithms
camPUs
The CAMPUS app is an all in one app, that keeps you updated with the latest events taking place in Panjab University
CSC2515-Introduction-to-Machine-Learning-Fall-2021
Assignments done as part of the course CSC2515 - Introduction to Machine Learning
Data-Structures
Some code for data structures
jupyterlab-dash
An Extension for the Interactive development of Dash apps in JupyterLab
Matlab
Programs to learn and implement matlab
NasaSpace
Disaster Prep is an app that aims to help users get prepared to face any natural calamity, ask for help in case of need and be ready before hand.
Python-Overview
A quick guide for learning python and relevant interview questions for practice