Press "Enter" to skip to content

Category: Data Science

Breeze: Mathematics In Scala

Nitin Aggarwal introduces the mathematics library behind Spark’s machine learning library, MLlib:

In simple terms, Breeze is a Scala library that extends the Scala collection library to provide support for vectors and matrices in addition to providing a whole bunch of functions that support their manipulation. We could safely compare Breeze to NumPy in Python terms. Breeze forms the foundation of MLlib—the Machine Learning library in Spark

Breeze comprises four libraries:

  • breeze-math: Numerics and Linear Algebra. Fast linear algebra backed by native libraries (via JBlas) where appropriate.

  • breeze-process: Tools for tokenizing, processing, and massaging data, especially textual data. Includes stemmers, tokenizers, and stop word filtering, among other features.

  • breeze-learn: Optimization and Machine Learning. Contains state-of-the-art routines for convex optimization, sampling distributions, several classifiers, and DSLs for Linear Programming and Belief Propagation.

  • breeze-viz: (Very alpha) Basic support for plotting, using JFreeChart.

Read on for samples and basic usage.

Comments closed

Data Science At A Small Tech Company

Julia Silge blogs about her first year as a data scientist at Stack Overflow:

In the fall I saw this post by Shanif Dhanani about being a data scientist at a small company, and it is entirely on point, the whole way through. So much of that post resonates with my own experience of being a data scientist at a small company. And yes, I do keep saying “small company”; Stack Overflow is likely smaller than you think it is, 250 or so employees in total. I am the second data scientist here, joining David Robinson who was the first data science hire, on a data team that is five in total.

I cannot emphasize enough how much of my day-to-day work is communicating, collaborating with others, and answering not-entirely-specified questions. Data science is highly technical work, but the value of my technical work would be much lower if I could not communicate what it means in clear and compelling ways. My definition of communication here is pretty broad, and includes speaking, writing, and data visualization.

If you’re interested in a career in data science, this is food for thought.

Comments closed

The Importance Of Model Interpretability

Ilknur Kaynar Kabul explains why it’s important that your data science models be interpretable:

Some machine learning models are simple and easy to understand. We know how changing the inputs will affect the predicted outcome and can make justification for each prediction. However, with the recent advances in machine learning and artificial intelligence, models have become very complex, including complex deep neural networks and ensembles of different models. We refer to these complex models as black box models.

Unfortunately, the complexity that gives extraordinary predictive abilities to black box models also makes them very difficult to understand and trust. The algorithms inside the black box models do not expose their secrets. They don’t, in general, provide a clear explanation of why they made a certain prediction. They just give us a probability, and they are opaque and hard to interpret. Sometimes there are thousands (even millions) of model parameters, there’s no one-to-one relationship between input features and parameters, and often combinations of multiple models using many parameters affect the prediction. Some of them are also data hungry. They need enormous amounts of data to achieve high accuracy. It’s hard to figure out what they learned from those data sets and which of those data points have more influence on the outcome than the others.

This post reminds me of a story I’d heard about a financial organization using neural networks to build accurate models, but then needing to decompose the models into complex decision trees to explain to auditors that they weren’t violating any laws in the process.

Comments closed

Hierarchical Clustering

Chaitanya Sagar explains hierarchical clustering with examples in R:

Hope now you have a better understanding of clustering algorithms than what you started with. We discussed about Divisive and Agglomerative clustering techniques and four linkage methods namely, Single, Complete, Average and Ward’s method. Next, we implemented the discussed techniques in R using a numeric dataset. Note that we didn’t have any categorical variable in the dataset we used. You need to treat the categorical variables in order to incorporate them into a clustering algorithm. Lastly, we discussed a couple of plots to visualise the clusters/groups formed. Note here that we have assumed value of ‘k’ (number of clusters) is known. However, this is not always the case. There are a number of heuristics and rules-of-thumb for picking number of clusters. A given heuristic will work better on some datasets than others. It’s best to take advantage of domain knowledge to help set the number of clusters, if that’s possible. Otherwise, try a variety of heuristics, and perhaps a few different values of k.

There’s a lot to pick out of this post, but you’re able to walk through it step by step.  H/T R-Bloggers

Comments closed

Matrix Transposition In T-SQL

Phil Factor has some fun transposing a matrix using T-SQL:

What I’m doing is simply converting the table into its JSON form, and then using this to create a table using the multi-row VALUES  syntax which paradoxically allows expressions. The expression I’m using is JSON_Value, which allows me do effectively dictate the source within the table, via that JSON Path expression, and the destination. As it is an expression, I can do all sorts of manipulation as well as a transpose.  I could, if I wanted, (in SQL 2017)provide that path parameter as a variable. This sort of technique can be used for several other reporting purposes, and it is well-worth experimenting with it because it is so versatile.

That is not at all what I would have thought up; very interesting approach.  I’d probably just be lazy and shell out to R Services.

Comments closed

Taking A Random Walk

Dan Goldstein describes the basics of Brownian motion:

I was sitting in a bagel shop on Saturday with my 9 year old daughter. We had brought along hexagonal graph paper and a six sided die. We decided that we would choose a hexagon in the middle of the page and then roll the die to determine a direction:

1 up (North)
2 diagonal to the upper right (Northeast)
3 diagonal to the lower right (Southeast)
4 down (South)
5 diagonal to the lower left (Southwest)
6 diagonal to the upper left (Northwest)

Our first roll was a six so we drew a line to the hexagon northwest of where we started. That was the first “step.”

After a few rolls we found ourselves coming back along a path we had gone down before. We decided to draw a second line close to the first in those cases.

We did this about 50 times. The results are pictured above, along with kid hands for scale.

Javi Fernandez-Lopez then shows how to generate an animated GIF displaying Brownian motion:

Last Monday we celebrated a “Scientific Marathon” at Royal Botanic Garden in Madrid, a kind of mini-conference to talk about our research. I was talking about the relation between fungal spore size and environmental variables such as temperature and precipitation. To make my presentation more friendly, I created a GIF to explain the Brownian Motion model. In evolutionary biology, we can use this model to simulate the random variation of a continuous trait through time. Under this model, we can notice how closer species tend to maintain closer trait values due to shared evolutionary history. You have a lot of information about Brownian Motion models in evolutionary biology everywhere!

Another place that this is useful is in describing stock market movements in the short run.

Comments closed

Estimating Used Car Prices

Kevin Jacobs wants to estimate the value of his car and shows how to set up a machine learning job to do this:

As you can see, I collected the brand (Peugeot 106), the type (1.0, 1.1, …), the color of the car (black, blue, …) the construction year of the car, the odometer of the car (which is the distance in kilometers (km) traveled with the car at this point in space and time), the ask price of the car (in Euro’s), the days until the MOT (Ministry of Transport test, a required periodical check-up of your car) and the horse power (HP) of the car. Feel free to use your own variables/units!

It’s an interesting example of how you can approach a real problem.

Comments closed

Outlier Detection In R

Giorgio Garziano has an introduction to outlier detection and intervention analysis using R:

Now, we implement a similar representation of the transient change outlier by taking advantage of the arimax() function within the TSA package. The arimax() function requires to specify some ARMA parameters, and that is done by capturing the seasonality as discussed in ref. [1]. Further, the transient change is specified by means of xtransf and transfer input parameters. The xtransf parameter is a matrix with each column containing a covariate that affects the time series response in terms of an ARMA filter of order (p,q). For our scenario, it provides a value equal to 1 at the outliers time index and zero at others. The transfer parameter is a list consisting of the ARMA orders for each transfer covariate. For our scenario, we specify an AR order equal to 1.

Check it out.

Comments closed

Predicting Restaurant Reservations With A Neural Net

Kevin Jacobs builds a simple neural net using Pandas and sklearn:

The first thing to notice is that our values are not normalized. The number of visitors is a number and gets larger and larger. To normalize it, we simply divide it by 100, since all numbers are below 1. The same holds for the lag. Most of the lags are lower than 30. Therefore, I will divide the lag size by 30.

Notice that there are many more approaches for normalizing the data! This is just a quick normalization on the data, but feel free to use your own normalization method. My normalization process is closely related to the MinMaxScalar normalization which can be found in sklearn (scikit-learn).

With just a few lines of Python code we can create a Multi-Layer Perceptron (MLP):

Click through for the code.

Comments closed

Using Keras To Predict Customer Churn

Matt Dancho has an example of building a neural net using Keras to predict customer churn:

Pro Tip: A quick test is to see if the log transformation increases the magnitude of the correlation between “TotalCharges” and “Churn”. We’ll use a few dplyr operations along with the corrr package to perform a quick correlation.

  • correlate(): Performs tidy correlations on numeric data

  • focus(): Similar to select(). Takes columns and focuses on only the rows/columns of importance.

  • fashion(): Makes the formatting aesthetically easier to read.

This is a very useful tutorial.

Comments closed