-
Anomalous sound pattern detection for industrial machine health monitoring: In this study the idea of converting audio
samples from different industrial machines such as pump, slider, fan, and valve to mel-spectrograms images using Short Time
Fourier Transform (STFT) is explored and studied in detail. The converted audio samples were then used to train convolutional neural
network (CNN) based models using “normal” mel-spectrogram images. The model performance was evaluated on a test data set having unseen
normal and anomaly samples. The performance of CNN architecture was compared with the baseline deep neural network trained using raw
audio samples. The results show that CNN based models perform better than models trained on the raw dataset (i.e., sound samples without spectrogram conversion).
978-3-031-68617-7
-
Time Series Visualization: This timeline tracks the theological roots of the Nazism and Apartheid eras and is designed to exhibit the
research work of Dr. Kimberly Vrudny, Associate Dean of Theology department at the University of St. Thomas. This work resulted in enhanced learning experience for
theology students and other audiences who keeps an interest in perceiving the occurrence of these events visually.
Nazism Apartheid timeline
-
Benchmarking Cassandra using YCSB: This work studies the popular Cassandra database in a distributed environment through an exhaustive
performance analysis. This analysis is conducted using data replication and data partitioning strategies leveraging the datasets of Yahoos
Cloud Serving Benchmark (YCSB).
Overview
-
Eleckart Ecommerce Case Study: Developed a market mix model to observe the impact of marketing variables in 2018 and recommended optimal budget
allocation for different marketing categories. Dataset consisted of 500K instances with 34 numerical and categorical features.
Overview
Master Dataset Aggregation
Exploratory Data Analysis
Market Mix Models - Linear, Multiplicative, Koyck
Market Mix Models - Distributed Lag
-
Uber Demand Supply Gap Analysis: Conducted exploratory data analysis, which identified the root cause behind the cancellation and the non-availability of cabs
and further recommended ways to improve the situation in Bengaluru, India. Dataset consisted of SK instances with 9 numerical and categorical features.
Overview
Exploratory Data Analysis
-
Bike Rental Analysis for Seoul city: Built regression models such as polynomial, decision tree, random forest, and ensemble model using scikit learn libraries, that predicted rental bike count for different weather conditions.
Dataset consisted of 9K instances with 14 numerical and categorical features.
Overview
Exploratory Data Analysis