I made a tool called MALSS (Machine Learning Support System) that supports machine learning in Python (PyPI/GitHub //github.com/canard0328/malss)). We would be grateful if you could give us various opinions.

I'm writing the continuation here. ・ Tool MALSS (basic) that supports machine learning in Python ・ Tool MALSS (application) that supports machine learning in Python

Purpose

The rich open source development environment has made it really easy to ** use machine learning **. However, because it is easy to use without studying, you may get stuck in an unexpected pitfall. MALSS was developed with the aim of automating the parts that beginners tend to get stuck in, as well as learning how to use machine learning. In addition, I think that even experts can contribute to efficiency by automating the introduction part of analysis.

Target

There are two applications of machine learning in business.

--Data analysis project -** System development project **

In data analysis projects, it is required to discover some knowledge from the data (data mining). On the other hand, in system development projects, a model that makes some predictions using unknown data is incorporated into the system and used. This tool is supposed to support the latter machine learning for ** system development projects **. This is because we believe that machine learning for system development projects has a larger part that can be patterned (automated) at the initial examination stage.

Also, as a classification of machine learning

-** Supervised learning ** --Unsupervised learning

there is. In supervised learning, data consists of a pair of input and output, and the output is a numerical value, regression that predicts it from the input, and the output is a category value, and which category is classified from the input. There is a classification / classification to predict. In unsupervised learning, data consists only of inputs, and this includes clustering, which categorizes inputs according to some criteria, and outlier detection. This tool is intended for the former ** supervised learning **. This is also because I think that supervised learning has a larger part that can be patterned.

Introduction

Dependencies

MALSS depends on the following packages. We do not automatically download the dependent packages at the time of installation, so please handle it manually. Anaconda is convenient because it contains everything.

• python (>= 2.7 or >= 3.4)
• numpy (>= 1.10.2)
• scipy (>= 0.16.1)
• scikit-learn (>= 0.17)
• matplotlib (>= 1.5.1)
• pandas (>= 0.14.1)
• jinja2 (>= 2.8)

Installation

From the command line pip install -U malss You can install it with.

Alternatively, download the source from PyPI or GitHub, unzip it, and put it in the extraction directory. Go, from the command line python setup.py install You can install it with.

Why python

This is because the machine learning part is left to scikit-learn.

in conclusion

I think there are still some points that have yet to be reached as a tool, so I would be grateful if you could give us various opinions. Next time would like to write about the specific usage.