diff --git a/Cargo.toml b/Cargo.toml
index 8b545fb6e..11240d9db 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -31,6 +31,7 @@ intel-mkl-static = ["blas", "ndarray-linalg", "intel-mkl-src/mkl-static-lp64-seq
 intel-mkl-system = ["blas", "ndarray-linalg", "intel-mkl-src/mkl-dynamic-lp64-seq"]
 
 blas = ["ndarray/blas"]
+polars = ["polars-core"]
 
 [dependencies]
 num-traits = "0.2"
@@ -40,7 +41,7 @@ ndarray = { version = "0.13", default-features = false, features = ["approx"] }
 ndarray-linalg = { version = "0.12.1", optional = true }
 
 [dependencies.intel-mkl-src]
-version = "0.6.0"
+path = "/home/lorenz/Downloads/intel-mkl-src/intel-mkl-src"
 default-features = false
 optional = true
 
@@ -56,6 +57,11 @@ optional = true
 default-features = false
 features = ["cblas"]
 
+[dependencies.polars-core]
+version = "0.12"
+optional = true
+default-features = false
+
 [dev-dependencies]
 ndarray-rand = "0.11"
 approx = { version = "0.3", default-features = false, features = ["std"] }
diff --git a/datasets/Cargo.toml b/datasets/Cargo.toml
index c2d87cd1a..eeddb805d 100644
--- a/datasets/Cargo.toml
+++ b/datasets/Cargo.toml
@@ -13,6 +13,8 @@ ndarray = { version = "0.13", default-features = false }
 ndarray-csv = "0.4"
 csv = "1.1"
 flate2 = "1.0"
+polars-core = { version = "0.12", default-features = false, optional = true }
+polars-io = { version = "0.12", default-features = false, optional = true }
 
 [dev-dependencies]
 approx = { version = "0.3", default-features = false, features = ["std"] }
@@ -23,3 +25,4 @@ diabetes = []
 iris = []
 winequality = []
 linnerud = []
+income = ["polars-core", "polars-io", "linfa/polars"]
diff --git a/datasets/data/income-test.tar.gz b/datasets/data/income-test.tar.gz
new file mode 100644
index 000000000..920c233e0
Binary files /dev/null and b/datasets/data/income-test.tar.gz differ
diff --git a/datasets/data/income-train.tar.gz b/datasets/data/income-train.tar.gz
new file mode 100644
index 000000000..42aa424e4
Binary files /dev/null and b/datasets/data/income-train.tar.gz differ
diff --git a/datasets/src/lib.rs b/datasets/src/lib.rs
index 8b9b64551..02235923a 100644
--- a/datasets/src/lib.rs
+++ b/datasets/src/lib.rs
@@ -39,6 +39,15 @@ use linfa::Dataset;
 use ndarray::prelude::*;
 use ndarray_csv::Array2Reader;
 
+#[cfg(feature = "income")]
+use linfa::dataset::Dataframe;
+#[cfg(feature = "income")]
+use std::io::{Cursor, Read};
+#[cfg(feature = "income")]
+use polars_core::frame::DataFrame;
+#[cfg(feature = "income")]
+use polars_io::{csv::CsvReader, SerReader};
+
 #[cfg(any(
     feature = "iris",
     feature = "diabetes",
@@ -58,6 +67,20 @@ fn array_from_buf(buf: &[u8]) -> Array2<f64> {
     reader.deserialize_array2_dynamic().unwrap()
 }
 
+#[cfg(feature = "income")]
+fn dataframe_from_buf(buf: &[u8]) -> DataFrame {
+    let mut file = GzDecoder::new(buf);
+    let mut buf: Vec<u8> = Vec::new();
+    file.read_to_end(&mut buf).unwrap();
+    let buf = Cursor::new(buf);
+
+    CsvReader::new(buf)
+        .infer_schema(None)
+        .has_header(false)
+        .finish()
+        .unwrap()
+}
+
 #[cfg(feature = "iris")]
 /// Read in the iris-flower dataset from dataset path.
 // The `.csv` data is two dimensional: Axis(0) denotes y-axis (rows), Axis(1) denotes x-axis (columns)
@@ -157,6 +180,14 @@ pub fn linnerud() -> Dataset<f64, f64> {
     Dataset::new(input_array, output_array).with_feature_names(feature_names)
 }
 
+#[cfg(feature = "income")]
+pub fn income() -> (Dataframe<bool>, Dataframe<bool>) {
+    let input_data = include_bytes!("../data/income-train.tar.gz");
+    let input_dataframe = dataframe_from_buf(&input_data[..]);
+
+    panic!("")
+}
+
 #[cfg(test)]
 mod tests {
     use super::*;
diff --git a/src/dataset/impl_records.rs b/src/dataset/impl_records.rs
index 49e9d62dd..97d74c5de 100644
--- a/src/dataset/impl_records.rs
+++ b/src/dataset/impl_records.rs
@@ -1,6 +1,9 @@
 use super::{DatasetBase, Float, Records};
 use ndarray::{ArrayBase, Axis, Data, Dimension};
 
+#[cfg(feature = "polars")]
+use polars_core::frame::DataFrame;
+
 /// Implement records for NdArrays
 impl<F: Float, S: Data<Elem = F>, I: Dimension> Records for ArrayBase<S, I> {
     type Elem = F;
@@ -52,3 +55,16 @@ impl<R: Records> Records for &R {
         (*self).nfeatures()
     }
 }
+
+#[cfg(feature = "polars")]
+impl Records for DataFrame {
+    type Elem = ();
+
+    fn nsamples(&self) -> usize {
+        self.shape().0
+    }
+
+    fn nfeatures(&self) -> usize {
+        self.shape().1
+    }
+}
diff --git a/src/dataset/mod.rs b/src/dataset/mod.rs
index b4096c6fd..446dd0321 100644
--- a/src/dataset/mod.rs
+++ b/src/dataset/mod.rs
@@ -8,6 +8,9 @@ use ndarray::{
 };
 use num_traits::{FromPrimitive, NumAssignOps, Signed};
 
+#[cfg(feature = "polars")]
+use polars_core::frame::DataFrame;
+
 use std::cmp::{Ordering, PartialOrd};
 use std::collections::{HashMap, HashSet};
 use std::hash::Hash;
@@ -147,6 +150,10 @@ pub type DatasetView<'a, D, T> = DatasetBase<ArrayView<'a, D, Ix2>, ArrayView<'a
 pub type DatasetPr<D, L> =
     DatasetBase<ArrayBase<OwnedRepr<D>, Ix2>, CountedTargets<L, ArrayBase<OwnedRepr<Pr>, Ix3>>>;
 
+#[cfg(feature = "polars")]
+/// Dataframe
+pub type Dataframe<T> = DatasetBase<DataFrame, ArrayBase<OwnedRepr<T>, Ix2>>;
+
 /// Record trait
 pub trait Records: Sized {
     type Elem;
diff --git a/src/prelude.rs b/src/prelude.rs
index 030686626..f1769c91f 100644
--- a/src/prelude.rs
+++ b/src/prelude.rs
@@ -13,6 +13,9 @@ pub use crate::traits::*;
 #[doc(no_inline)]
 pub use crate::dataset::{AsTargets, Dataset, DatasetBase, DatasetView, Float, Records};
 
+#[cfg(feature = "polars")]
+pub use crate::dataset::Dataframe;
+
 #[doc(no_inline)]
 pub use crate::metrics_classification::{BinaryClassification, ConfusionMatrix, ToConfusionMatrix};