amir-mjafari Goto Github PK

aadp-estimation

We introduced ML methods based on site characteristics (e.g., nearby land use), weather conditions, date, and short-term counts (STC) to estimate Daily Pedestrian Traffic. The data was available for 94 stations (intersections) in Arizona, US, with an average of 223 continuous counting days at each intersection for passing pedestrians. A Neural Network Model trained by Keras library had an R-squared of 0.82, more explanatory than the Linear Regression Model with an R-squared of 0.71.

amir-mjafari

Config files for my GitHub profile.

classification_oxford_102_flowers

ecommerce-classification

I used Keras library to train a Convolutional Neural Network Model to classify 21627 noisy images of e-commerce products into 27 categories. Due to the availability of corresponding text descriptions and categorical features, I implemented a Transformer Model (pre-trained BERT) with PyTorch library to process the text data. The ensemble of the two mentioned models achieved 96 percent accuracy.

image-processing

Image Manipulation, Edge Detection, Skeletons, and Feature Extraction with MATLAB.

join_us_app

Simple Join-US Web-based Application

netflix

Using Response Surface Methodology (RSM) to minimize users' browsing times in Netflix's movie suggestion system by adjusting the system layout configuration such as Tile size, Preview length and etc.

regex_experiment

Using REGEX to extract information (i.e., Hospital Name, Address, Postal Code, etc.) from tabular data in PDF format for a gap Analysis. The next phase is using the Data frame (Instead of the messy table in PDF format) to find unconsidered costumers and reaching them out.

resecon703

Topics in Advanced Econometrics (ResEcon 703). University of Massachusetts Amherst. Taught by Matt Woerman

shopify_challenge_ds_2022

Shopify Data Science Internship Challenge Summer 2022

slot_filling_and_intent_detection_of_slu

slot filling, intent detection, joint training, ATIS & SNIPS datasets, the Facebook’s multilingual dataset, MIT corpus, E-commerce Shopping Assistant (ECSA) dataset, CoNLL2003 NER, ELMo, BERT, XLNet

time_serries_analysis

Conducting a SARIMA model on average hourly travel speeds data (1 month data) for an specific route (Prediction of travel times in the Origin-Destination Level)