mickydowns Goto Github PK

central_limit

courses

Course materials for the Data Science Specialization: https://www.coursera.org/specialization/jhudatascience/1

datasciencecoursera

repo for johns hopkins / coursera data science course

datasharing

The Leek group guide to data sharing

deep-cnn-sat-image-time-series

Deep CNN and RNN for Satellite Image Time Series

deep-keras-lstm-momo-transformer

Modifications to code from the paper Trading with the Momentum Transformer: An Intelligent and Interpretable Architecture (https://arxiv.org/pdf/2112.08534.pdf).

deep-keras-music-net

AI musical instrument practice aid

deep-theano-rnn-lstm-car

Pack of Drones: Layered reinforcement learning (Q-learning w/ RNN) for complex "hunt" behaviors

deep_tf_notminst

two sets of (tensorflow) neural nets: 1. convnets for notMINST image recognition, 2. RNNs for natural language processing

estimate_mpg

fin_bank_failure

predict california bank failures during 2008 recession using linear regression and discriminant methods

fin_bond_option_pricing

estimate bond prices using un/constrained optimization, spot/instantaneous rates and vasicek and cev models

fin_equity_option_pricing

pricing european puts/calls and up-and-out puts using closed form and monte carlo simulation (cev process, euler and millstein discretization)

fin_hft_analytics

analysis of high frequency quote and trade data for F ang GM. finds a variety of "seasonal" patterns using vwap and duration measures.

fin_integrative_models

This project uses a three-tiered strategy to increasae returns from the randomley selected portfolio of stocks. Tier one is a baseline, long-only, buy-hold portfolio. Tier two begins with the same portfolio, but trades component stocks daily, taking long and short positions as determined by a set of technical rules. Tier three trades component stocks monthly based on portfolio rules that view stock movements relative to the entire market.

fin_portfolio_volatility

builds stock portfolios based on: 1. price momentum strategies as described by Jegadeesh and Titman (2001), 2. tripartite momentum as described by Lewellen (2002), and 3. volatility quantile as proposed by Han, Yang and Zhou (2013). reports sharpe, treynor, sortino, calmar, information, and capm alpha metrics and benchmarks against buy-and-hold strategy.

fin_pricing_asian_options

pricing asian equity options used closed form (Turnbull, Levy and Geometric Average) and monte carlo simulation. employ a range of variance reduction techniques including antithetic variates, common random numbers, control variates, conditional monte carlo, stratified sampling and importance sampling.

fin_technical_rules

uni (e.g., momentum, RSI), bi (i.e., Anatolyev) and trivariate (i.e., cointegration) technical trading rules for JPM, WFC, SPY

fin_time_series_fundamentals

implements basic time series analysis including: estimating IBM intra-day trade durations using Epanechnikov and GCV, fitting a MARS model for SP500 options and futures prices based on moneyness and days-to-exp, performing ADF unit-root test for a multivariate rate time series, finding and estimating cointegration vectors, and regressing COFI on selected rates.

fin_variance_reduction

compares variance in monte carlo put and call option price forecasts with and without antithetic variate variance reduction technique (assumes discretized price paths based on normal and/or uniform distributions)

fin_weather_damage

analysis of weather events from NOAA database to classify events with greatest: 1. harm to population health, 2. economic consequences

household-power-consumption

Plotting Assignment 1 for Exploratory Data Analysis

iot_particulate_matter

analysis of fine particulate mater (pm2.5) produced from U.S. emission sources from 1999 to 2008.

jhs-ml-project

mine_algorithms

code implementing data mining algos (svm, boosting, random forest knn, etc.) on sonar data set. good for comparing strengths/weaknesses of each algo.

mine_hh_incomes

digs into performance differential across tree strategies including random forest and gradient boosting machines when predicting household incomes.

mine_ncaa_rankings

using linear and logistic methods w/ sparse data structure, this analysis: 1. predicts team wins/losses, then 2. NCAA ranking.

mine_pca_v_pls

code contrasts partial least squares regression (plsr) vs principal components regression (pcr) for predicting gender based on body measurements

mine_regression

regression on boston data set including non-linear transforms, regression splines, MARS

mine_search_results

early work mining keyword and web page features to yield a workable search engine algorithm