Chris

Modeling Revenue Data – Part IV: ARIMA

ARIMA is known for its effectiveness in modeling time series data. In this section, we will try to fit the data using an ARIMA model. Theoretically, the parameters, (p, q), of an ARIMA(p, d, q) is selected using partial autocorrelation and autocorrelation coefficients. The parameter of d implies the number of times we take differencesContinue reading “Modeling Revenue Data – Part IV: ARIMA”

Bitcoin Price Prediction

PLEASE DO NOT USE THE MODEL IN THIS ARTICLE FOR TRADING OR USE IT AT YOUR OWN RISK. LIVE MARKETS ARE VERY COMPLICATED AND THE MODEL IN THIS ARTICLE WILL NOT GUARANTEE PROFITS. In this article, let’s try to utilize LSTM techniques to predict the price of bitcoin. Unlike what we have done for theContinue reading “Bitcoin Price Prediction”

Modeling Revenue Data – Part III: LSTM

In this part, let’s build an LSTM model for the sales revenue data. Long short temporary model is developed on the basis of Recurrent Neural Network to improve its performance in NLP by retaining some information from “long” past data. Due to its capability of processing sequence data, LSTM can also be used to modelContinue reading “Modeling Revenue Data – Part III: LSTM”

Modeling Revenue Data – Part II: Linear Regression Approach

Using the same data set as in Part I, let’s explore some methods to apply linear regression for time-series data. Our data looks like this: To model the data using linear regression, we need to convert date and time into features and scale the revenue into the appropriate range. And now our dataset looks likeContinue reading “Modeling Revenue Data – Part II: Linear Regression Approach”

Modeling Revenue Data – Part I: EDA

Exploration of the dataset This article shows the modeling process for a sales/revenue dataset. The dataset contains 2,935,846 records of sales data from Jan 01, 2013 to Oct 31, 2015 with 60 vendors and 22170 items. Brief description of the dataset is shown below. To make this dataset easier to work with, I will re-formatContinue reading “Modeling Revenue Data – Part I: EDA”

Multivariate Linear Regression with Gradient Descent

In this article, I will try to extend the material from univariate linear regression into multivariate linear regression (mLR). In mLR, n features are collected for each observation, and is now also a vector of dimension n+1 where is the intercept, or the coefficient for an arbitrary feature of x with all values equal toContinue reading “Multivariate Linear Regression with Gradient Descent”

Gradient Descent for Univariate Linear Regression

The purpose of this article is to realize gradient descent algorithm for univariate linear regression in Python. We first briefly show the mathematics of the algorithm and then the code in Python to realize gradient descent method. The cost function that will be used is mean square error (MSE). Let’s define some notations that willContinue reading “Gradient Descent for Univariate Linear Regression”