Tag: ai

MNIST Example

In this module, you will learn how to use Google Colab while using the well known MNIST example

Tags:

2 minute read

Project: Detection of Autism Spectrum Disorder with a Facial Image using Artificial Intelligence

This project uses artificial intelligence to explore the possibility of using a facial image analysis to detect Autism in children. Early detection and diagnosis of Autism, along with treatment, is needed to minimize some of the difficulties that people with Autism encounter. Autism is usually diagnosed by a specialist through various Autism screening methods. This can be an expensive and complex process. Many children that display signs of Autism go undiagnosed because there families lack the expenses needed to pay for Autism screening and diagnosing. The development of a potential inexpensive, but accurate way to detect Autism in children is necessary for low-income families. In this project, a Convolutional Neural Network (CNN) is utilized, along with a dataset obtained from Kaggle. This dataset consists of collected images of male and female, autistic and non-autistic children between the ages of two to fourteen years old. These images are used to train and test the CNN model. When one of the images are received by the model and importance is assigned to various features in the image, an output variable (autistic or non-autistic) is received.

Tags:

11 minute read

Report: Dentronics: Classifying Dental Implant Systems by using Automated Deep Learning

Artificial intelligence is a branch of computer science that focuses on building and programming machines to think like humans and mimic their actions. The proper concept definition of this term cannot be achieved simply by applying a mathematical, engineering, or logical approach but requires an approach that is linked to a deep cognitive scientific inquiry. The use of machine-based learning is constantly evolving the dental and medical field to assist with medical decision making process. In addition to diagnosis of visually confirmed dental caries and impacted teeth, studies applying machine learning based on artificial neural networks to dental treatment through analysis of dental magnetic resonance imaging, computed tomography, and cephalometric radiography are actively underway, and some visible results are emerging at a rapid pace for commercialization. Researchers have found deep convolutional neural networks to have a future place in the dental field when it comes to classification of dental implants using radiographic images.

Tags:

9 minute read

Report: Aquatic Animals Classification Using AI

Marine animals play an important role in the ecosystem. ‘Aquatic animals play an important role in nutrient cycles because they store a large proportion of ecosystem nutrients in their tissues, transport nutrients farther than other aquatic animals and excrete nutrients in dissolved forms that are readily available to primary producers’ (Vanni MJ 1) Fish images are captured by scuba divers, tourist, or underwater submarines. different angles of fishes image can be very difficult to get because of the constant movement of the fish. In addition to getting the right angles, the images of marine animals are usually low-quality because of the water. Underwater cameras that is required for a good quality image can be expensive. Using AI could potentially increase the marine population by the help of classification by testing the usage of machine learning using the images obtained from the aquarium combined with advanced technology. We collect 164 fish images data from Georgia acquarium to look at the different movements.

Tags:

4 minute read

Project: Hand Tracking with AI

In this project we study the ability of an AI to recognize letters from the American Sign Language (ASL) alphabet. We use a Convolutional Neural Network and apply it to a dataset of hands in different positionings showing the letters ‘a’, ‘b’, and ‘c’ in ASL. With this we build a model to recognize the letter and output the letter it predicts.

Tags:

8 minute read

Aquatic Toxicity Analysis with the aid of Autonomous Surface Vehicle (ASV)

With the passage of time, human activities have created and contributed much to the aggrandizing problems of various forms of environmental pollution. Massive amounts of industrial effluents and agricultural waste wash-offs, that often comprise pesticides and other forms of agricultural chemicals, find their way to fresh water bodies, to lakes, and eventually to the oceanic systems. Such events start producing a gradual increase in the toxicity levels of marine ecosystems thereby perturbing the natural balance of such water-bodies. In this endeavor, an attempt will be made to analyze the various water quality metrics (viz. temperature, pH, dissolved-oxygen level, and conductivity) that are measured with the help of autonomous surface vehicles (ASV). The collected data will undergo big data analysis tasks so as to find the general trend of values for the water quality of the given region. These obtained values will then be compared with sample water quality values obtained from neighboring sources of water for ascertaining if these sample values exhibit aberration from the established values that were found earlier from the big data analysis tasks for water-quality standards. In the event, the sample data popints significantly deviate from the standard values established earlier, it can then be successfully concluded that the aquatic system in question, from which the water sample was sourced from, has been degraded and may no longer be utilized for any form of human usage, such as being used for drinking water purposes.

Tags:

31 minute read

How Big Data has Affected Statistics in Baseball

The purpose of this report is to highlight how the inception of big data in baseball has changed the way baseball is played and how it affects the choices managers make before, during, and after a game. It was found that big data analytics can allow baseball teams to make more sound and intelligent decisions when making calls during games and signing contracts with free agent and rookie players. The significance of this project and what was found was that teams that adopt the moneyball mentality would be able to perform at much higher levels than before with a much lower budget than other teams. The main conclusion from the report was that the use of data analytics in baseball is a fairly new idea, but if implemented on a larger scale than only a couple of teams, it could greatly change the way baseball is played from a managerial standpoint.

Tags:

10 minute read

Predictive Model For Pitches Thrown By Major League Baseball Pitchers

The topic of this review is how big data analysis is used in a predictive model for classifying what pitches are going to be thrown next. Baseball is a pitcher’s game, as they can control the tempo. Pitchers have to decide what type of pitch they want to throw to the batter based on how their statistics compare to that of the batters. They need to know what the batter struggles to hit against, and where in the strike zone they struggle the most. With the introduction of technology into sports, data scientists are sliding headfirst into Major League Baseball. And with the introduction of Statcast in 2015, The MLB has been looking at different ways to use technology in the game. In 2020 alone, the MLB introduce several different types of technologies to keep the fans engaged with the games while not being able to attend them [^3]. In this paper, we will be exploring a predictive model to determine pitches thrown by each pitcher in the MLB. We will be reviewing several predictive models to understand how this can be done with the use of big data.

Tags:

9 minute read

Big Data Analytics in the National Basketball Association

The National Basketball Association and the deciding factors in understanding how the game should be played in terms of coaching styles, positions of players, and understanding the efficiencies of shooting certain shots is something that is prevalent in why analytics is used. Analytics is a topic space within basketball that has been growing and emerging as something that can make a big difference in the outcomes of gameplay. With the small analytic departments that have been incorporated within teams, results have already started coming in with the teams that use the analytics showing more advantages and dominance over opponents who don’t. We will analyze positions on the court of players and how big data and analytics can further take those positions and their game statistics and transform them into useful strategies against opponents.

Tags:

18 minute read

Big Data in E-Commerce

The topic of my report is big data in e-commerce. E-commerce is a big part of todays society. During the shopping online, the recommend commodities are fitter and fitter for my liking and willingness to buy. This is the merit of big data. Big data use my purchase history and browsing history to analyze my liking and recommend the goods for me.

Tags:

13 minute read

Big Data Analytics in Brazilian E-Commerce

As the world begins to utilize online service and stores at greater capacity it becomes a greater priority to increase the efficiency of the various processes that are required for online stores to work effectively. By analyzing the data the comes from online purchases, a better understanding can be formed about what is needed and where as well as the quantity. This data should also allow for us to better predict what orders will be needed at future times so shortages can be avoided.

Tags:

14 minute read

Rank Forecasting in Car Racing

The IndyCar Series is the premier level of open-wheel racing in North America. Computing System and Data analytics is critical to the game, both in improving the performance of the team to make it faster and in helping the race control to make it safer. IndyCar ranking prediction is a practical application of time series problems. We will use the LSTM model to analyze the state of the car, and then predict the future ranking of the car. Rank forecasting in car racing is a challenging problem, which is featured with highly complex global dependency among the cars, with uncertainty resulted from existing exogenous factors, and as a sparse data problem. Existing methods, including statistical models, machine learning regression models, and several state-of-the-art deep forecasting models all perform not well on this problem. In this project, we apply deep learning methods to racing telemetry data. And compare deep learning with traditional statistical methods (SVM, XGBoost).

Tags:

9 minute read

Change of Internet Capabilities Throughout the World

In 2050 the United Nations is projecting that 90% of the world will have access to the internet. With the recent pandemic and the shift to most things being online we see how desperate people need internet to be able to do everyday tasks. The internet is a valuable utility and more people are getting access to it every day. We also are seeing more data is being sent over the internet with more than 24,000 Gigabytes being uploaded and processed per second across the entire internet. In this report we look at the progression of the internet and how it has changed over the years.

Tags:

25 minute read

Project: Chat Bots in Customer Service

Automated customer service is a rising phenomon for buisnesses with an online presence. As customer service bots advance in complication of problems they can handle one concern about the altered customer experiece is how the information is conveyed. Using customer support data tweets on twitter this project runs sentiment analysis on it customer tweets and then train a convolutional neural network to examine if conversation tone can be detected early in the conversation.

Tags:

7 minute read

COVID-19 Analysis

By the end of 2019, healthcare across the world started to see a new type of Flu and they called it Coronavirus or Covid-19. This new type of Flu developed across the world and it appeared there is no one treatment could be used to treat it yet, scientists found different treatments that apply to different age ranges. In this project, We will try to work on comparison analysis between USA and China on number of new cases and new deaths and trying to find factors played big roles in this spread.

Tags:

7 minute read

Analyzing the Relationship of Cryptocurrencies with Foriegn Exchange Rates and Global Stock Market Indices

The project involves analyzing the relationships of various cryptocurrencies with Foreign Exchange Rates and Stock Market Indices. Apart from analyzing the relationships, the objective of the project is also to estimate the trend of the cryptocurrencies based on Foreign Exchange Rates and Stock Market Indices. We will be using historical data of 6 different cryptocurrencies, 25 Stock Market Indices and 22 Foreign Exchange Rates for this project. The project will use various machine learning tools for analysis. The project also uses a fully connected deep neural network for prediction and estimation. Apart from analysis and prediction of prices of cryptocurrencies, the project also involves building its own database and giving access to the database using a prototype API. The historical data and recent predictions can be accessed through the public API.

Tags:

15 minute read

Project: Deep Learning in Drug Discovery

Machine learning has been a mainstay in drug discovery for decades. Artificial neural networks have been used in computational approaches to drug discovery since the 1990s. Under traditional approaches, emphasis in drug discovery was placed on understanding chemical molecular fingerprints, in order to predict biological activity. More recently however, deep learning approaches have been adopted instead of computational methods. This paper outlines work conducted in predicting drug molecular activity, using deep learning approaches.

Tags:

12 minute read

Big Data Application in E-commerce

As a result of the last twenty year’s Internet development globally, the E-commerce industry is getting stronger and stronger. While customers enjoyed their convenient online purchase environment, E-commerce sees the potential for the data and information customers left during their online shopping process. One fundamental usage for this information is to perform a Recommendation Strategy to give customers potential products they would also like to purchase. This report will build a User-Based Collaborative Filtering strategy to provide customer recommendation products based on the database of previous customer purchase records. This report will start with an overview of the background and explain the dataset it chose Amazon Review Data. After that, each step for the code and step made in a corresponding file Big_tata_Application_in_E_commense.ipynb will be illustrated, and the User-Based Collaborative Filtering strategy will be presented step by step.

Tags:

17 minute read

Residential Power Usage Prediction

We are living in a technology-driven world. Innovations make human life easier. As science advances, the usage of electrical and electronic gadgets are leaping. This leads to the shoot up of power consumption. Weather plays an important role in power usage. Even the outbreak of Covid-19 has impacted daily power utilization. Similarly, many factors influence the use of electricity-driven appliances at homes. Monitoring these factors and consolidating them will result in a humungous amount of data. But analyzing this data will help to keep track of power consumption. This system provides a prediction of usage of electric power at residences in the future and will enable people to plan ahead of time and not be surprised by the monthly electricity bill.

Tags:

14 minute read

Big Data Applications in the Gaming Industry

Gaming is one of the fastest growing aspects of the modern entertainment industry. It’s a rapidly evolving market, where trends can change in a near instant, meaning that companies need to be ready for near anything when making decisions that may impact development times, targets and milestones. Companies need to be able to see market trends as they happen, not post factum, which frequently means predicting things based off of freshly incoming data. Big data is also used for development of the games themselves, allowing for new experiences and capabilities. It’s a relatively new use for big data, but as AI capabilities in games are developed further this is becoming a very important method of providing more immersive experiences. Last use case that will be talked about, is monetization in games, as big data has also found a use there as well.

Tags:

20 minute read

Project: Forecasting Natural Gas Demand/Supply

Natural Gas(NG) is one of the valuable ones among the other energy resources. It is used as a heating source for homes and businesses through city gas companies and utilized as a raw material for power plants to generate electricity. Through this, it can be seen that various purposes of NG demand arise in the different fields. In addition, it is essential to identify accurate demand for NG as there is growing volatility in energy demand depending on the direction of the government’s environmental policy. This project focuses on building the model of forecasting the NG demand and supply amount of South Korea, which relies on imports for much of its energy sources. Datasets for training include various fields such as weather and prices of other energy resources, which are open-source. Also, those are trained by using deep learning methods such as the multi-layer perceptron(MLP) with long short-term memory(LSTM), using Tensorflow. In addition, a combination of the dataset from various factors is created by using pandas for training scenario-wise, and the results are compared by changing the variables and analyzed by different viewpoints.

Tags:

13 minute read

Big Data on Gesture Recognition and Machine Learning

Since our technology is more and more advanced as time goes by, traditional human-computer interaction has become increasingly difficult to meet people’s demands. In this digital era, people need faster and more efficient methods to obtain information and data. Traditional and single input and output devices are not fast and convenient enough, it also requires users to learn their own methods of use, which is extremely inefficient and completely a waste of time. Therefore, artificial intelligence comes out, and its rise has followed the changeover times, and it satisfied people’s needs. At the same time, gesture is one of the most important way for human to deliver information. It is simple, efficient, convenient, and universally acceptable. Therefore, gesture recognition has become an emerging field in intelligent human-computer interaction field, with great potential and future.

Tags:

29 minute read

Big Data in the Healthcare Industry

Healthcare is an organized provision of medical practices provided to individuals or a community. Over centuries the application of innovative healthcare has been needed increasingly as humans expand their life span and become more aware of better preventative care practices. The application of Big Data within the industry of Healthcare is of the utmost importance in order to quantify the effects of wide scale efficient and safe solutions. Pharmaceutical and Bio Data Research companies can use big data to intake large facets of patient record data and use this collected data to iterate how preventative care can be implemented before diseases actually present themselves in stages that are beyond the point of potential recovery. Data collected in laboratory settings and statistics collected from medical and state institutions of healthcare facilitate time, money, and life saving initiatives as deep learning can in certain instances perform better than the average doctor at detecting malignant cells. Big data within healthcare has proven great results for the advancement and diverse application of informed reasoning towards medical solutions.

Tags:

18 minute read

Analysis of Various Machine Learning Classification Techniques in Detecting Heart Disease

As cardiovascular diseases are the number 1 cause of death in the United States, the study of the factors and early detection and treatment could improve quality of life and lifespans. From investigating how the variety of factors related to cardiovascular health relate to a general trend, it has resulted in general guidelines to reduce the risk of experiencing a cardiovascular disease. However, this is a rudimentary way of preventative care that allows for those who do not fall into these risk categories to fall through. By applying machine learning, one could develop a flexible solution to actively monitor, find trends, and flag patients at risk to be treated immediately. Solving not only the risk categories but has the potential to be expanded to annual checkup data revolutionizing health care.

Tags:

18 minute read

Predicting Hotel Reservation Cancellation Rates

As a result of the Covid-19 pandemic all segments of the travel industry face financial struggle. The lodging segment, in particular, has had the financial records scrutinized revealing a glaring problem. Since the beginning of 2019, the lodging segment has seen reservation cancellation rates near 40%. At the directive of business and marketing experts, hotels have previously attempted to solve the problem through an increased focus on reservation retention, flexible booking policies, and targeted marketing. These attempts did not produce results, and continue to leave rooms un-rented which is detrimental to the bottom line. This document will explain the creation and testing of a novel process to combat the rising cancellation rate. By analyzing reservation data from a nationwide hotel chain, it is hoped that an algorithm may be developed capable of predicting the likeliness that a traveler is to cancel a reservation. The resulting algorithm will be evaluated for accuracy. If the resulting algorithm has a satisfactory accuracy, it would make clear to the hotel industry that the use of big data is key to solving this problem.

Tags:

13 minute read

Analysis of Future of Buffalo Breeds and Milk Production Growth in India

Water buffalo (Bubalus bubalis) is also called Domestic Water Buffalo or Asian Water Buffalo. It is large bovid originating in Indian subcontinent, Southeast Asia, and China and today found in other regions of world - Europe, Australia, North America, South America and some African countries. There are two extant types recognized based on morphological and behavioral criteria: 1. River Buffalo - Mostly found in Indian subcontinent and further west to the Balkans, Egypt, and Italy. 2. Swamp Buffalo - Found from west of Assam through Southeast Asia to the Yangtze valley of China in the east. India is the largest milk producer and consumer compared to other countries in the world and stands unique in terms of the largest share of milk being produced coming from buffaloes. The aim of this academic project is to study the livestock census data of buffalo breeds in India and their milk production using Empirical Benchmarking analysis method at state level. Looking at the small sample of data, our analysis indicates that we have been seeing increasing trends in past few years in livestock and milk production but there are considerable opportunities to increase production using combined interventions.

Tags:

16 minute read

Music Mood Classification

Music analysis on an individual level is incredibly subjective. A particular song can leave polarizing impressions on the emotions of its listener. One person may find a sense of calm in a piece, while another feels energy. In this study we examine the audio and lyrical features of popular songs in order to find relationships in a song’s lyrics, audio features, and its valence. We take advantage of the audio data provided by Spotify for each song in their massive library, as well as lyrical data from popular music news and lyrics site, Genius.

Tags:

18 minute read

NBA Performance and Injury

Sports Medicine will be a $7.2 billion dollar industry by 2025. The NBA has a vested interest in predicting performance of players as they return from injury. The authors evaluated datasets available to the public within the 2010 decade to build machine and deep learning models to expect results. The team utilized Gradient Based Regressor, Light GBM, and Keras Deep Learning models. The results showed that the coefficient of determination for the deep learning model was approximately 98.5%. The team recommends future work to predicting individual player performance utilizing the Keras model.

Tags:

23 minute read

NFL Regular Season Skilled Position Player Performance as a Predictor of Playoff Appearance Overtime

The present research investigates the value of in-game performance metrics for NFL skill position players (i.e., Quarterback, Wide Receiver, Tight End, Running Back and Full Back) in predicting post-season qualification. Utilizing nflscrapR-data that collects all regular season in-game performance metrics between 2009-2018, we are able to analyze the value of each of these in-game metrics by including them in a regression model that explores each variables strength in predicting post-season qualification. We also explore a comparative analysis between two time periods in the NFL (2009-2011 vs 2016-2018) to see if there is a shift in the critical metrics that predict post-season qualification for NFL teams. Theoretically, this could help inform the debate as to whether there has been a shift in the style of play in the NFL across the previous decade and where those changes may be taking place according to the data. Implications and future research are discussed.

Tags:

19 minute read

Project: Training A Vehicle Using Camera Feed from Vehicle Simulation

Deep Learning has become the main form of machine learning that has been used to train, test, and gather data for self-driving cars. The CARLA simulator has been developed from the ground up so that reasearchers who normally do not have the capital to generate their own data for self-driving vehicles can do so to fit their spcific model. CARLA provides many tools that can simulate many scenarios that an autonomous vehicle would run into. The benefit of CARLA is that it can simulate scenarios that may be too dangerous for a real vehicle to perform, such as a full self-driving car in a heavly populated area. CARLA has the backing of many companies who lead industry like Toyota who invested $100,000 dollars in 2018. This project uses the CARLA simulator to visualize how a real camera system based self-driving car sees obstacles and objects.

Tags:

9 minute read

Project: Structural Protein Sequences Classification

The goal of this project is to predict the family of a protein based on the amino acid sequence of the protein. The structure and function of a protein are determined by the amino acid sequence that composes it. In the protein structure data set, each protein is classified according to its function. Categories include: HYDROLASE, OXYGEN TRANSPORT, VIRUS, SIGNALING PROTEIN, etc. dozens of kinds. In this project, we will use nucleic acid sequences to predict the type of protein. Although there are already protein search engines such as BLAST that can directly query the known protein families. But for unknown proteins, it is still important to use deep learning algorithms to predict their functions. Protein classification is a simpler problem than protein structure prediction. The latter requires the complete spatial structure of the protein, and the required deep learning model is extremely complex.

Tags:

7 minute read

How Big Data Can Eliminate Racial Bias and Structural Discrimination

Healthcare is utilizing Big Data to to assist in creating systems that can be used to detect health risks, implement preventative care, and provide an overall better experience for patients. However, there are fundmental issues that exist in the creation and implementation of these systems. Medical algorithms and efforts in precision medicine often neglect the structural inequalities that already exist for minorities accessing healthcare and therefore perpetuate bias in the healthcare industry. The author examines current applications of these concepts, how they are affecting minority communities in the United States, and discusses improvements in order to achieve more equitable care in the industry.

Tags:

22 minute read

Sentiment Analysis and Visualization using a US-election dataset for the 2020 Election

Sentiment analysis is an evaluation of the opinion of the speaker, writer, or other subjects about some topic. We are going to use the US-elections dataset and combining the tweets of people’s opinions for leading presidential candidates. We have various datasets from Kaggle and combining tweets and NY times datasets, by combining all data prediction will be derived.

Tags:

17 minute read

Estimating Soil Moisture Content Using Weather Data

As the world is gripped with finding solutions to problems such as food and water shortages, the study of agriculture could improve where we stand with both of these problems. By integrating weather and sensor data, a model could be created to estimate soil moisture based on weather data that is easily accessible. While some farmers could afford to have many moisture sensors and monitor them, many would not have the funds or resources to keep track of the soil moisture long term. A solution would be to allow farmers to contract out a limited study of their land using sensors and then this model would be able to predict soil moistures from weather data. This collection of data, and predictions could be used on their own or as a part of a larger agricultural solution.

Tags:

21 minute read

Big Data in Sports Game Predictions and How It is Used in Sports Gambling

Big data in sports is being used more and more as technology advances and this has a very big impact, especially when it comes to sports gambling. Sports gambling has been around for a while and it is gaining popularity with it being legalized in more places across the world. It is a very lucrative industry and the bookmakers use everything they can to make sure the overall odds are in their favor so they can reduce the risk of paying out to the betters and ensure a steady return. Sports statistics and data is more important than ever for bookmakers to come up with the odds they put out to the public. Odds are no longer just determined by expert analyzers for a specific sport. The compilation of odds uses a lot of historical data about team and player performance and looks at the most intricate details in order to ensure accuracy. Bookmakers spend a lot of money to employ the best statisticians and the best algorithms. There are also many companies that solely focus on sports data analysis, who often work with bookmakers around the world. On the other hand, big data for sports game analysis is also used by gamblers to gain a competitive edge. Many different algorithms have been created by researchers and gamblers to try to beat the bookmakers, some more successful than others. Oftentimes these not only involve examining sports data, but also analysing data from different bookmakers odds in order to determine the best bets to place. Overall, big data is very important in this field and this research paper aims to show the various techniques that are used by different stakeholders.

Tags:

21 minute read

Analyzing LSTM Performance on Predicting the Stock Market for Multiple Time Steps

Predicting the stock market has been an attractive field of research for a long time because it promises big wealth for anyone who can find the secret. For a long time, traders around the world have been relying on fundamental analysis and technical analysis to predict the market. Now with the advancement of big data, some financial institutions are beginning to predict the market by creating a model of the market using machine learning. While some researches produce promising results, most of them are directed at predicting the next day’s market behavior. In this study, we created an LSTM model to predict the market for multiple time frames. We then analyzed the performance of the model for some different time periods. From our observations, LSTM is good at predicting 30 time steps ahead, but the RMSE became larger as the time frame gets longer.

Tags:

16 minute read

Stock Price Reactions to Earnings Announcements

On average the US stock market sees a total of $192 billion a day trading hands. Massive companies, hedge funds, and other high level institutions use the markets to capitalize on companies potential and growth over time. The authors used Financial Modeling Prep to gather over 20 years of historical stock data and earnings calls to understand better what happens during company’s earnings annoucements. The results showed that over a large sample size of companies, identifying a strong coorelation was rather difficult, yet companies with strong price trend tendencies were more predictable to beat earnings expectations.

Tags:

17 minute read

Project: Stock level prediction

This project includes a deep learning model for stock prediction. It uses LSTM, RNN which is the standart for time series prediction. It seems to be the right approach. The author really loved this project since he loves stocks. He invests often, and is also in love with tech, so he finds ways to combine both of them. Most existing models for stock prediction do not include the volume, and Rishabh intendede to use that as an input, but it did not go exactly as planned.

Tags:

6 minute read

Review of Text-to-Voice Synthesis Technologies

The paper is about the most popular and most successful voice synthesis methods in the recent 5 years. Area of examples that would be explored in order to produce such a review paper would consist of both academic research papers and examples real world successful applications. For each specific example examined, its dataset, theory/model, training algorithms, and the purpose and use for that specific method/technology would be examined and reviewed. Overall, the paper will compare the similarities and differences between these methods and explore how big data enabled these new voice-synthesis technologies. And last, the changes these technologies will bring to our world in the future is discussed and both positive and negatives implications are explored in depth. This paper is meant to be informative to the both general audience and professionals about the how voice-synthesizing techniques has been transformed by big data, most important developments in the academic research of this field, and how these technologies are adopted to create innovation and value. But also to explain the logic and other technicalities behind these algorithms created by academia and applied to real world purposes. Codes and datasets of voices will be supplemented as for the purpose of demonstrations of these technologies in working.

Tags:

16 minute read

Analysis of Financial Markets based on President Trump's Tweets

President Trump has utilized the social media platform Twitter as a way to convey his message to the American people. The tweets he has published during his presidency cover a vast array of topics and issues from MAGA rallies to impeachment. This analysis investigates the relationship of the NASDAQ and the sentiment of President Trump’s tweets during key events in his presidency. NASDAQ data was gathered though Yahoo Finance’s API while President Trump’s tweets were gathered from Kaggle. The results observed show that during certain events, a correlation emerges of the NASDAQ data and the sentiment of President Trump’s tweets.

Tags:

13 minute read

Trending Youtube Videos Analysis

The internet has created a revolution for how people connect, understand topics, and consume information. Today, the consumption of the media is easier than ever. Going onto the internet and finding interesting content takes less than a minute to do. In the already growing industry of amateur or professional video production, Youtube is one of many go-to platforms for viewers and creators to collide. Social media creates an avenue for Youtubers to help promote their videos and reach a wider audience. For hours on end, viewers can watch nearly any type of content uploaded onto the site. However, it is harder for video creators to make an interesting video any person can enjoy than a viewer to find one of those videos. In the congested mass of videos, how can a Youtuber create a unique identity allowing their videos to go viral? This report will address this issue by creating a prediction of how Youtube popularizes a video and a solution to help a video go viral.

Tags:

16 minute read

Review of the Use of Wearables in Personalized Medicine

Wearable devices offer an abundant source of data on wearer activity and health metrics. Smartphones and smartwatches have become increasingly ubiquitous, and provide high-quality motion sensor data. This research attempts to classify movement types, including running, walking, sitting, standing, and going up and down stairs, to establish the practicality of sharing this raw data with healthcare workers. It also addresses the existing research regarding the use of wearable data in clinical settings and discusses shortcomings in making this data available.

Tags:

10 minute read

Project: Identifying Agricultural Weeds with CNN

Weed identification is an important component of agriculture, and can affect the way farmers utilize herbicide. When unable to locate weeds in a large field, farmers are forced to blanket utilize herbicide for weed control. However, this method is bad for the environment, as the herbicide can leech into the water, and bad for the farmer, because they then must pay for far more fertilizer than they really need to control weeds. This project utilizes images from the Aarhus University [^1] dataset to train a CNN to identify images of 12 species of plants. To better simulate actual rows of crops, a subset of the images for testing will be arranged in a list representing a crop row, with weeds being distributed in known locations. Then, the AI is tested on the row, and should be able to determine where in the row the weeds are located.

Tags:

9 minute read

Detect and classify pathologies in chest X-rays using PyTorch library

Chest X-rays reveal many diseases. Early detection of disease often improves the survival chance for Patients. It is one of the important tools for Radiologists to detect and identify underlying health conditions. However, they are two major drawbacks. First, it takes time to analyze a radiograph. Second, Radiologists make errors. Whether it is an error in diagnosis or delay in diagnosis, both outcomes result in a loss of life. With the technological advances in AI, Deep Learning models address these drawbacks. The Deep Learning models analyze the X-rays like a Radiologist and accurately predict much better than the Radiologists. In our project, first, we develop a Deep Learning model and train our model to use the labels for Atelectasis, Cardiomegaly, Consolidation, Edema, and Pleural Effusion that corresponds to 5 different diseases, respectively. Second, we test our model’s performance: how well our model predicts the diseases. Finally, we visualize our model’s performance using the AUC-ROC curve.

Tags:

16 minute read

All Tags