Connect ML

Account

Learn More

Login

Create Account

Logout

Machine Learning

Analyse, Visualise and Predict, with one click

Upload data and run for free

Explore Machine Learning methods to uncover insights in your data

{{ algorithms[i-1] }}
{{(accuracies[i-1]*100).toFixed(1)}}%

Run Machine Learning Algorithms

Compare different Classification and Regression algorithms. Choose the best to make predictions

View Statistical Properties

Spearman, Pearson Correlation

Chi Square Tests

Point Biseral Correlation

Find the most important features

Mutual Information

Random Forest Rank

15 Algorithms

Automatic Data Cleaning

Advanced Visualisations

Human Readable Reports

Statistical Analysis

Variable Analysis

Free

For Datasets up to 50 MB

API

Coming soon

Sign Up

Creating account

Sign up with Google

Sign up with Twitter

Sign up with Microsoft

Algorithms

Visualize

Prediction

{{ best }}

{{ best }} is the best performing algorithm for this dataset.
This suggests your dataset does not have complex interactions between independent variables,
and that variables share linear relationships with the prediction variable.
Linear Regression has the benefit of being easy to interpret, and is a relatively simple model to reason about mathametically.
For more information see
OLS
{{ best }}

{{ best }} is the best performing algorithm for this dataset.
This suggests your dataset does not have complex interactions between independent variables,
and that variables share linear relationships with the prediction variable.
Lasso Regression performs both variable selection and regularization, this means the model is relatively more
robust to overfitting and can handle messy data.
Lasso Regression has the benefit of being easy to interpret, and is a relatively simple model to reason about mathametically.
For more information see
Lasso Regression
{{ best }}

{{ best }} is the best performing algorithm for this dataset.
This suggests your dataset does not have complex interactions between independent variables,
and that variables share linear relationships with the prediction variable.
Ridge Regression uses regularization, meaning variables with low prediction value and highly correlated
independent variables are given a lower weighting for prediction. This has the benefit of avoiding overfitting,
this is particularly valuable when the dataset does not follow all of the assumptions for standard regression to
give unbiased results. For more information see
Ridge Regression
{{ best }}

{{ best }} is the best performing algorithm for this dataset.
This suggests your dataset does not have complex interactions between independent variables,
and that variables share linear relationships with the prediction variable.
Ridge Regression uses regularization, meaning variables with low prediction value and highly correlated
independent variables are given a lower weighting for prediction. This has the benefit of avoiding overfitting,
this is particularly valuable when the dataset does not follow all of the assumptions for standard regression to
give unbiased results. For more information see
Bayesian Ridge Regression
{{ best }}

{{ best }} is the best performing algorithm for this dataset.
This suggests your dataset does not have complex interactions between independent variables,
and that variables share linear relationships with the prediction variable. For more information see
Linear Discriminant Analysis
{{ best }}

{{ best }} is the best performing algorithm for this dataset.
Naive Bayes works by estimating which of possible outputs of a prediction variable is most likely,
given the conditional probability estimates of the inputs. Naive Bayes does not take into account
interactions between independent variables. This suggests your dataset does not contain interactions,
and the relationships between inputs and the output do not require any transformation. Naive Bayes has
the beneif of being easy to interpret and reason about. For more information see
Naive Bayes
{{ best }}

{{ best }} is the best performing algorithm for this dataset.
K-Nearest-Neighbours excels in domains where there is limited training instances and/or many class labels.
It works well on data with linear and non linear features, and can capture variable interactions.
It's main advantage is that it is non parametric, meaning it does not
learn any parameters, this is useful for avoiding overfitting on smaller datasets.
K Nearest Neighbours works by selecting the average of the closest K neighbours for any
prediction. The above algorithm uses the closest 3 neighbours for any point.
For more information see
K-Nearest-Neighbours
{{ best }}

{{ best }} is the best performing algorithm for this dataset.
Random Forests excel in datasets where there are complex interactions between independent variables,
and have the advantage of being robust to overfitting. Random Forests use ensembles of Decision Trees,
which use gini impurity as a method for selecting features. This embeded feature selection ability makes
Random Forests robust on datasets with many redundant or low powered features.
For more information see
Random Forest
{{ best }}

{{ best }} is the best performing algorithm for this dataset.
Ada Boost excels on datasets where there are complex interactions between independent variables.
Ada Boost uses an ensemble of Decision Trees, which use gini impurity as a method for selecting features.
This embeded feature selection ability makes Ada Boosting robust on on datasets with many redundant or low powered features.
For more information see
Ada Boost
{{ best }}

{{ best }} is the best performing algorithm for this dataset.
Gradient Boosters excel in datasets where there are complex interactions between independent variables,
and have the advantage of being robust to overfitting. Gradient Boosters use ensembles of Decision Trees,
which use gini impurity as a method for selecting features.
This embeded feature selection ability makes Gradient Boosters robust on on datasets with many redundant or low powered features.
For more information see
Gradient Boosting
{{ best }}

{{ best }} is the best performing algorithm for this dataset.
Neural Networks are universal function approximators, making them an incredibly flexible Machine Learning method.
Neural Networks have the benefit of capturing interactions non linear interactions between independent variables.
The above Network uses a single hidden layer, feed forward network for prediction. This suggests your data has
complex non linear interactions. The downside of Neural Networks are that they are relatively more difficult to
interpret, and can overfit without enough training data.
For more information see
Neural Network
{{ best }}

{{ best }} is the best performing algorithm for this dataset.
This suggests variables in your dataset have non linear relationships with the predictor variable.
SVMs can efficiently perform a non-linear classification using the kernel trick, mapping inputs into high-dimensional feature spaces.
For more information see
Support Vector Machine
{{ best }}

{{ best }} is the best performing algorithm for this dataset.
This suggests the prediction variable in your dataset can be estimated using a simple model.
Decision Trees work particularly well with datasets with many categorical variables, and can capture
interactions among variables. Decision Trees can work on messy datasets, since they place a lower
weighting on low powered or redundant features. For more information see
Decision Tree
{{ best }}

{{ best }} is the best performing algorithm for this dataset.
This suggests the dataset can be modelled using a relatively simple algorithm, where capturing interactions among
independent variables is not necessary for predicint the output variable. For more information see
Logistic Regression
You haven't uploaded any datasets to predict with yet

Upload new prediction dataset

{{ uploadText }}

Upload new prediction dataset

{{ uploadText }}

Run Selected Algorithms

{{ a }}

{{ a }}

Algorithms

Visualize

Prediction

Feature Importance

Variable Analysis

Algorithms

Data Graph

Dimensionality Reduction

Feature Importance ({{ rankType }})

Mutual Information

Random Forest Rank

Mutual Information

Mutual Information quantifies the amount of information shared between two variables.
Mutual Information can capture complex dependencies, which may not be detected using metrics such as linear correlation.
Mutual Information can also compare information between categorical, binary and continuous variables.
We use a non parametric method based on entropy estimation from the K-Nearest-Neighbours. Scores are normalized
between 0 and 1, where 0 indicates no information is shared between the variables, and 1 indicates indentical information.
For more information see Mutual Information

Random Forest Rank

The Random Forest feature ranking method has the advantage of capturing interactions between independent variables, unlike Mutual
Information or Correlation methods. It can also compare continious, discrete and boolean variables.
Random Forests are an enemble of randomly sampled Decision Trees, and use a weighting of these
to rank which features are most important. Decision Trees use the Gini impurity measure to quantify
a feature's importance. For more information see
Random Forests and
Decision Tree
Export Data

Loading

Variable Analyses ({{visualData.variableNames[selectedVariable]}})

Select Variable

Type:

{{ visualData.variables[visualData.variableNames[selectedVariable]].type }}

Mean:

{{ visualData.variables[visualData.variableNames[selectedVariable]].mean }}

Min:

{{ visualData.variables[visualData.variableNames[selectedVariable]].min }}

Max:

{{ visualData.variables[visualData.variableNames[selectedVariable]].max }}

Variance:

{{ visualData.variables[visualData.variableNames[selectedVariable]].variance }}

# Missing values

{{ visualData.variables[visualData.variableNames[selectedVariable]].numberMissing }}

# Unique Values

{{ visualData.variables[visualData.variableNames[selectedVariable]].numberUnique }}

Relationship with Predictor

Mutual Information

{{ selectedFS }}

Point-Biserial Correlation

{{ visualData.stats.pointBiserial[varNames[selectedVariable]][predictorName].correlation }}

P-Value: {{ visualData.stats.pointBiserial[varNames[selectedVariable]][predictorName].pvalue.toFixed(2) }}
- {{ visualData.stats.pointBiserial[varNames[selectedVariable]][predictorName].summary }}

Spearman Correlation

{{ visualData.stats.spearman[varNames[selectedVariable]][predictorName].correlation }}

P-Value: {{ visualData.stats.spearman[varNames[selectedVariable]][predictorName].pvalue.toFixed(2) }}
- {{ visualData.stats.spearman[varNames[selectedVariable]][predictorName].significance }}

Pearson Correlation

{{ visualData.stats.pearson[varNames[selectedVariable]][predictorName].coefficient }}

P-Value: {{ visualData.stats.pearson[varNames[selectedVariable]][predictorName].pvalue.toFixed(2) }}
- {{ visualData.stats.pearson[varNames[selectedVariable]][predictorName].significance }}

Chi Square Test

{{ visualData.stats.chi2[varNames[selectedVariable]][predictorName].coefficient }}

P-Value: {{ visualData.stats.chi2[varNames[selectedVariable]][predictorName].pvalue.toFixed(2) }}
- {{ visualData.stats.chi2[varNames[selectedVariable]][predictorName].significance }}

Comparing {{ visualData.variableNames[selectedVariable] }} with {{ visualData.variableNames[selectedCompareVariable] }}

Change Comparison Variable

Statistical Information

Method

Coefficient

P-Value

Significance

Spearman Correlation

{{ visualData.stats.spearman[varNames[selectedVariable]][varNames[selectedCompareVariable]].correlation }}

{{ visualData.stats.spearman[varNames[selectedVariable]][varNames[selectedCompareVariable]].pvalue }}

{{ visualData.stats.spearman[varNames[selectedVariable]][varNames[selectedCompareVariable]].significance }} |
{{ visualData.stats.spearman[varNames[selectedVariable]][varNames[selectedCompareVariable]].corr }} Linear Correlation

Pearson Correlation

{{ visualData.stats.pearson[varNames[selectedVariable]][varNames[selectedCompareVariable]].coefficient }}

{{ visualData.stats.pearson[varNames[selectedVariable]][varNames[selectedCompareVariable]].pvalue }}

{{ visualData.stats.pearson[varNames[selectedVariable]][varNames[selectedCompareVariable]].significance }} |
{{ visualData.stats.pearson[varNames[selectedVariable]][varNames[selectedCompareVariable]].rel }}

Chi Squared Test

{{ visualData.stats.chi2[varNames[selectedVariable]][varNames[selectedCompareVariable]].coefficient }}

{{ visualData.stats.chi2[varNames[selectedVariable]][varNames[selectedCompareVariable]].pvalue }}

{{ visualData.stats.chi2[varNames[selectedVariable]][varNames[selectedCompareVariable]].significance }}

Point Biserial Correlation

{{ visualData.stats.pointBiserial[varNames[selectedVariable]][varNames[selectedCompareVariable]].correlation }}

{{ visualData.stats.pointBiserial[varNames[selectedVariable]][varNames[selectedCompareVariable]].pvalue }}

{{ visualData.stats.pointBiserial[varNames[selectedVariable]][varNames[selectedCompareVariable]].summary }}

Mutual Information

{{ visualData.stats.mutualInformation[varNames[selectedVariable]][varNames[selectedCompareVariable]] }}

Exclude Variables

Prediction Variable

Algorithm Type

Exclude Variables

{{v.name}}

{{v.name}}

{{ predictionVariable }}

{{v}}

{{ mode }}

Classification

Regression

Cancel

Confirm settings and run