Cyprian Fusi's Projects
Applying AI for Sentiment Analysis with IMDB Dataset using tensorflow 2.0
The goal is to analyze data to help our developers understand what type of apps are likely to attract more users
Hypothesis testing and feature engineering in machine learning
The most simplified explanation of the Central Limit Theorem
Complete Statistical Hypothesis Test using real-world data is a blueprint for hypothesis testing! It covers almost all the hypothesis tests commonly used.
Config files for my GitHub profile.
National Tutoring Programme (NTP) and Accident and Emergency (A&E) Performance. With recommendations to UK Department for Education of 10 Local Authorities where NTP should be intensified and a response to UK Secretary of State for Health regarding a 76% target in A&E performance.
With this model: the amount of backlog would be reduced significantly, the amount of staff needed to do the job would be reduced drastically, the processing time would be shortened significantly and more cases of fraudulent transactions would be tracked down in a given amount of data processed - more than 40% increase in efficiency!
Getting started!
Image Classification using the German Traffic Sign Recognition Benchmark (GTSRB) using tensorflow2.0
An 87% efficient Spam Filter implemented from scratch using Naive Bayes Algorithm.
Validation RMSLE obtained: 0.21163 which is less than the RMSLE score (0.22909) that won the Kaggle Competition.
One of the most controversial issues in the U.S. educational system is the efficacy of standardized tests and whether they're unfair to certain groups. We could correlate SAT scores with factors like race, gender, income, and more.
Up to 90% accuracy with just 5 features using KNN algorithm and PCA for feature engineering. The dataset contained less than 1000 observations. The model's accuracy could be improved using more observations, further hyperparameter optimization and feature engineering
Analysing the Impact of Russian Tweets on the 2016 US Presidential Elections
Crawling and Scraping IP Addresses from Wikipedia Revision History Pages
LSTM and GRU models for sentiment predictions
Analysis of the various movie streaming platforms such as Netflix, Amazon Prime, Disney and Hulu
In this exercise we will train a network similar to the LeNet5 using Tensorflow2.0 and use it for inference.
With recommendation to the UK government to halt all mandatory testing! Tests should only be conducted on patients as part of diagnosis and treatment. This is because with low prevalence of the disease most positive test results are false positives. This is due to irreducible error in the test.
Using complex SQL queries to answer specific business questions. Notably using multiple named subqueries, views to extract data from a database to address specific problems.
This repo in due time shall contain notebooks with exercises demonstrating various methods of validating ML models
World Happiness Report for 2019 with strange and unexpected results for Sub-Sahara African Countries! But it's data speaking...
No reasonable amount of tickets would increase your chances of winning in a lottery. Stop the addiction!