CV

Research Interests

• Machine Learning
• Natural Language Processing
• Social science
• Data Mining
• Human Computer Interactions
• Sentiment and Belief
• Deep Learning
• Time Series Forecasting

Education

• Amirkabir University of Technology (Tehran Polytechnic), Tehran, Iran
  • B.Sc, Computer Engineering, Specialization package: Artificial Intelligence 2016 - 2020(expected)
  • GPA: (3.58/4)
• Allameh Tabatabi High School, Tehran, Iran
  • High School Diploma in Mathematics and Physics, 2012 - 2016
  • GPA: 19.37 / 20

Publications

• MM.Abdollahpour, P.Farinneya and E.Hajizadeh, “A new transformer-based hybrid model for forecasting energy market prices”, Energy ,Elsevier. (under review) abstract Talks
• P.Farinneya, MM.Abdollahpour and S.Hamidian “Active transfer learning in low resource Rumour detection”, The ACM Transactions on Asian and Low-Resource Language Information Processing (TALLIP). (Preparing)

Research and Academic experience

• Rumour detection, reply stance and source veracity (rumoureval 2019)
  • Researched on rumoureval 2019 dataset of rumorous tweets. This research consists of two tasks of source tweet veracity and reply tweets stance, experimented on Bert and Roberta document representation and different models from deep neural nets to ada boost to hybrid models and etc, also experimented on the effectiveness of using meta data and considering different time frames in training. This research is for finding a generalized and robust method for rumour detection and verification.
  • Supervisor: Doctor Hamidian
• Time series forecasting with hybrid Transformer-based models
  • A paper about how transformer-based models, which use self attention can be used to predict time series data, In this case Brent oil price, better than other methods such as LSTM , MLP , SVR and etc. With transformer model out performing all the other methods I developed a hybrid-ensemble model with all multiple different models to improve the performance of the final prediction model .
  • Supervisor: Doctor Hajizadeh
• Active Transfer Leanring for text classification
  • Currently I’m researching on developing an Active Transfer learning method for text classification. My motivation for this project is to train models that aren’t data hungry to train but can produce high accuracy results. In this research I experiment with finding the write combination of :
    • text representation ( Bag of Word, tf_idf, Glove, Bert, Tweet Bert, …)
    • model ( MLP with different architectures, SVM, Random forest, ada boost, …)
    • active learning query strategy ( Least certainty, Core set, Margin based, …) The goal is to train model that needs the least labeled data to hit a certain accuracy.
• Stock movement with social media information
  • Research on different approaches for stock price prediction and movement based on tweets about stocks and stock market movement and price. From a dataset crawled from twitter on different stocks, user scoring and social media aggregated sentiment and signals were obtained to feed into various models( Transformers, GRU , LSTMs and etc) to predict market movement and stock prices.
  • Supervisor: Doctor Nickabadi
• WNUT 2020 tasks 2 and 3
  • Task 2 working on developing a system that determines if a tweet (meta data crawled from twitter) is informative and contains information about COVID-19. Task 3 working on extracting COVID-19 events from Twitter and Using CovidBERT NER as feature input to improve provided baselines.
• Information Retrieval (Academic Course)
  • Implementing an online search engine for 150k persion news articles, which includes preprocessing (tokenization, stemming, stopwords) and indexing, vectorizing, scoring, clustering(knn, kmeans) and classifying news subjects.
• Data Mining (Academic Course)
  • project: Implementing linear regression and a classification challenge including preprocessing a large amount of data and employing algorithms like naive bayes. Implementing clustering problems by Kmeans and DBSCAN algorithms and using techniques like Elbow to identify best parameters. Implementing Nueral Networks to be trained and predict.
• Artificial Intelligence (Academic Course)
  • Implementing a foreground background detector for live webcam program that has dynamic snow falling on a persons head.
• Computational intelligence (Academic Course) *Implementing Evolution strategies and RBF to classify data.
• Algorithm Design (Academic Course)
  • project: Graph community detection using label propagation (lpam).

skills

• Programming :
  • Python, Java, MATLAB, JavaScript, Assembly,C++
• Frameworks and Tools :
  • Keras, TensorFlow, Pandas, Sklearn, PyTorch ,Matplotlib, Opencv ,huggingface,nltk
• Web Development :
  • JavaScript, vue js, HTML, CSS
• Operation Systems :
  • OSX , Linux , windows
• Database :
  • SQL

Teaching Experience

*Principles of Compiler Design - Fall 2020

• prepared home work assignments and helped students on developing a compiler *Data Structures - Fall 2020
• managed online quizzes to evaluate students on different data structures and algorithms, prepared home work assignments coding questions *Electrical circuits - Fall 2018
• graded home work assignments and quizzes *Algorithm Design - Spring 2018
• graded home work assignments and helped on coding assignments

Notable courses

Data structure: 20/20                                Algorithm Design: 19.25/20          
Information Retrieval: 17.96/2                       Data Mining: PASS*    
Computational Intelligence : 18/20                   Engineering Statistics: 17.2/20
Applied linear Algebra: 16.43/20                     Principles of  Programming: 19/20
Entrepreneurship and Business Planning: 20/20        Introduction to Bioinformatics: PASS*
Principles of Database Design: 17.3/20               Operating Systems: 17.8/20
* COVID-19 Semester Grades were binary due to COVID-19