English / Japanese

Introduction

Version 3.1.0 contains a number of new features, along with significant improvements to existing features.

For existing users, please note the following changes:

New features and major improvements are listed below. You can refer to this document by Help > About R AnalyticFlow at any time.

What's New and Improved

Analysis Features

Features for predictive/statistical analysis, data processing and visualization are improved.

The following operations have been newly added:

Features/options has been added to the operations below:

Here operation is what was called analysis module on documents. Although the term has been changed, there is no difference in usage. Likewise custom module is now called custom operation.

Other Improvements

Changes

Please note the following changes as indicated above:

Details of Improvements

Predictive Analysis

Features for predictive analysis have been improved. You can create multiple models at once and compare the performance of these models.

In Build Predictive Models operation select List and Models tab appears.

Specify model settings on the Models tab. You can compare models with varying parameters even with the same method.

It is also possible to evaluate predictive performance of models. It can be accessed on Evaluation tab in Build Predictive Models operation.

Evaluation can be output as predictive performance index (accuracy; ratio of correct prediction in this example), confusion matrix or charts as gains chart.

Accuracy Confusion Matrix Gains Chart

The same sort of evaluation can be done using Evaluation operation, by providing a data.frmae object including both predicted and actual values. Please note that charts as gains chart cannot be output if the predicted values are categories, not probabilities.

Statistical Analysis

Features for statistical analysis have been added, including hypothesis testing (t-test, Wilcoxon test, etc.) and multivariate analysis (principal component analysis and cluster analysis)

In hypothesis testing you can switch methods by just selecting option. Multiple comparison methods are also available.

Principal component analysis and cluster analysis can be used in similar ways. You can apply them only to numerical variables, and standardization is available before applying these methods. There is also an option to add generated cluster IDs to the original data as a new column.

Charts

New setting parameters have been added to X-Y plot, enabling more precise drawing controls. More practical, and beautiful charts can be drawn with it.

A new operation to draw box plot has also been added.

Data Processing

Features for data processing have been improved, and operations on data types and missing values have been added.

You can specify a data type (class) for each column. To set them on reading a data file, data types can be chosen as an option of Read Text File operation.

Data types of an existing data frame object can be changed with Set Data Types operation.

Missing Values operation provides the following processing regarding missing values:

These simple replacement and removal and mainly designed for predictive analysis. Especially when "missingness" itself has an important meaning in prediction, missing flags can improve predictive accuracy.

By pressing the Auto Generate Entries button processing entries for variables with missing values are automatically generated. If there are same processing entries for different variables, they can be grouped by Group button .

Missing value handling is one of important topics in Statistics, and many effective methods have been developed as multiple imputation methods (available on R with mice package). In the situation as you are interested in the significance of coefficients of a model, consider using such methods instead of these simple replacement or removal.

Operations on Selected Objects

Operations on selected objects can more easily be accessed.

To apply an operation applicable to a selected object, right-click on the object and click on Operation, or push the button on the viewer. You can also drag and drop the operation directly from here.

Editor

Text files and R script files can be edited. Select a file and press the Edit button to use the editor.

R Script File operation has also been improved in relation to this feature. This operation executes an R script file by source function, and now the specified R script file can be edited directly on the workflow.

Files being edited can be run even when they are unsaved (as temporary files), so they can be used in a similar way to standard R script nodes.

Project Import/Export

Analysis projects can be exchanged as files.

To output a project to a file, select Project > Export... from the menu. Select files and folders you want to output, and press the Export button to output them to a zip file.

The file size may be large, but it is also possible to include metadata (cached R objects and backups).

To extract and use the exported project, select Project > Import... from the menu, or you can import a project on the dialog shown by pushing button.

By importing the file, a zip file is expanded and can be used as a project.

Find/Replace

It is now possible to search and replace character strings contained in the node's input interface. It makes it easier to change the object names at once, for example.

If you want to search the generated scripts themselves as before, set the Script Search checkbox on.

Copy/Export R Scripts

The feature to output R scripts corresponding to workflow or node has been improved.

You can include node names in addition to comments, and the result can be confirmed in the preview before outputting. Scripts can be written to a file, or copied to the clipboard.

R as Another Process

Previously R was running on the same process as this software, but now R is running as another process. It brings us the following advantages:

To restart R when this software is running, click on Restart R from the console pane.

It can be used in the case as no response can be obtained from R. Please be aware that all existing R objects will be removed, and the state of R will be initialized.