SUPPORT_VECTOR_MACHINE
Download Flojoy Studio to try this app
Train a support vector machine (SVM) model for classification tasks. It takes two dataframes of label and feature from labelled training data and a dataframe of unlabelled input data. Params: kernel : 'linear' | 'poly' | 'rbf' | 'sigmoid' | 'precomputed' Specifies the kernel type to be used in the algorithm.
For detailed information about kernel types:
https://scikit-learn.org/stable/modules/svm.html#kernel-functions Returns: out : DataFrame The predictions for the input data.
Python Code
from flojoy import flojoy, DataFrame, Matrix
import pandas as pd
from sklearn import svm, preprocessing
from typing import Literal
@flojoy(deps={"scikit-learn": "1.2.2"})
def SUPPORT_VECTOR_MACHINE(
train_feature: DataFrame | Matrix,
train_label: DataFrame | Matrix,
input_data: DataFrame | Matrix,
kernel: Literal["linear", "poly", "rbf", "sigmoid", "precomputed"] = "linear",
) -> DataFrame:
"""Train a support vector machine (SVM) model for classification tasks.
It takes two dataframes of label and feature from labelled training data and a dataframe of unlabelled input data.
Parameters
----------
kernel : 'linear' | 'poly' | 'rbf' | 'sigmoid' | 'precomputed'
Specifies the kernel type to be used in the algorithm.
For detailed information about kernel types:
https://scikit-learn.org/stable/modules/svm.html#kernel-functions
Returns
-------
DataFrame
The predictions for the input data.
"""
le = preprocessing.LabelEncoder()
if isinstance(train_feature, DataFrame):
train = train_feature.m.to_numpy()
col = train_label.m.to_numpy()
target_name = train_label.m.columns.values[0]
else:
train = train_feature.m
col = train_label.m
target_name = "target"
X = train
Y = le.fit_transform(col)
clf = svm.SVC(kernel=kernel)
clf.fit(X, Y)
if isinstance(input_data, DataFrame):
input_arr = input_data.m.to_numpy()
else:
input_arr = input_data.m
prediction = clf.predict(input_arr)
prediction = le.inverse_transform(prediction)
prediction = pd.DataFrame({target_name: prediction})
return DataFrame(df=prediction)
Example App
Having problems with this example app? Join our Discord community and we will help you out!
In this example, the SUPPORT_VECTOR_MACHINE
is passed the iris dataset, split into two parts. The training data contains 120 labels examples, while the input dataset contains 30 samples with the labels removed.
This data is read from disk with two READ_CSV
nodes, and the output predictions made by the classifier are visualised in a TABLE
.