NPS - Exploratory analysis in R

August 28, 2017 in R, NLP

NPS analysis What is net promorter score (NPS)? Net Promoter Score or NPS is a customer loyalty metric and was developed by Fred Reichheld and it asks respondents to answer a single question. How likely are you to recommend this product? The respondents are asked to score between 0 and 10. 10 being “most likely” to recommend and 0 being “least likely”. An additional optional question is asked about why they picked this score and the response to that is usually a text comment.

Forecasting in R - Philadelphia crime data

October 26, 2016 in R

Exploring crime in Philadelphia This is a large and intersting dataset and has data points stretching back over 10 years. Several explorations have pointed out that crime seems to be seasonal and I wanted to explore this with a time series. Assuming that seasonal trends might repeat themselves, I am exploring this using the forecast package and using linear regression to predict trends. suppressPackageStartupMessages({ library(data.table) library(forecast) library(knitr) }) Data size and structure.

Access log parsing with Spark and schema extraction from query string

August 3, 2016 in Spark, Python

I needed to parse server logs and create Spark DataFrames to query information from the query string parameters. My naive version kept throwing errors about mismatched number of fields in schema and those in the row being queried. It turns out I was dealing with over 350 different query string params across the logs. This could change over time and there was no way I was going to add these programmatically by hand.

Revenue forecast for seasonal business

May 9, 2016 in R

Revenue forecasting with linear methods. This post highlights linear regression techniques on time series data. We have some weekly sales data from a small business that is expected to have seasonal trends. A bike rental business in a major tourist city. Let’s dive into the data. Load up the required packages. suppressPackageStartupMessages({ library(forecast) }) Load the data, we have two columns with weekly revenue numbers from two different sources. Online and Phone; these correspond to sales online and via phone call to the business.

On Marketplaces and platforms

February 17, 2014 in Platforms

Marketplaces are all the rage these days, in truth, marketplaces have been around since ancient times. A marketplace usually exists to connect sellers of goods or services with buyers. The marketplace itself usually benefits by selling space to sellers or by taking a cut of the transaction. This post is focused on service marketplaces that are connecting and democratizing human capital in a way that has never been possible. Things like vetting, payments, reviews and support have been researched and A/B tested to an art.

Finding with find

April 11, 2013 in Unix

Find is a really versatile utility that can be used to enumerate files of different types, narrow the list by file types, dates, sizes, access times and a whole list of expressions. The output can be formatted with various switches to be csv. My goal was to list the sizes and access times for all video files in the system. I knew that there was over 3 TB of files but not how recently these were accessed/played.

NPS - Exploratory analysis in R

Forecasting in R - Philadelphia crime data

Access log parsing with Spark and schema extraction from query string

Revenue forecast for seasonal business

On Marketplaces and platforms

Finding with find

Nitin Ahuja