https://pypi.org/project/mltronsAutoDataPrep/
Data is the most important element for data analysis. Real world data is unclean with a lot of spelling errors, missing values, formatting issues, skewness, no encoding or aggregation which makes it the most time-consuming & cumbersome task for analysts & scientists. As most of the scientists spend time around 80% of their time cleaning & preparing data, therefore we’re introducing dptron to make that process extremely easier and faster!
Dptron is an in-memory platform built for distributed & scalable data cleaning & preparation. DPtron is written in Python and is built on PySpark to deal with large amounts of data seamlessly. It uses an implementation of machine learning and deep learning algorithms to perform important data cleaning & preparation steps automatically. Dptron is extensible so that developers, analysts & scientists can streamline the process of data cleaning & preparation for better decision making while becoming more productive.
Decision making is better & easier if the data is clean otherwise it’s garbage-in and garbage-out.
- Supports connection with AWS S3
- Supports upto 10TB of data size
- Treats spelling mistakes and other inconsistencies in URLs
- Detects & treats skewness in data
- Feature engineering for time variable
- Treats & fills NULL values by using deep learning (next iteration)
- Treats spelling mistakes and other inconsistencies in other variables (next iteration)
Open up your terminal and install Java8 required for pySpark:
brew cask install adoptopenjdk/openjdk/adoptopenjdk8**After installing Java8, set it as your default Java version:
/usr/libexec/java_home -V**This will output thefollowing:
Matching Java Virtual Machines (3):
1.8.0_05, x86_64: "Java SE 8" /Library/Java/JavaVirtualMachines/jdk1.8.0_05.jdk/Contents/Home
1.6.0_65-b14-462, x86_64: "Java SE 6" /System/Library/Java/JavaVirtualMachines/1.6.0.jdk/Contents/Home
1.6.0_65-b14-462, i386: "Java SE 6" /System/Library/Java/JavaVirtualMachines/1.6.0.jdk/Contents/Home
/Library/Java/JavaVirtualMachines/jdk1.8.0_05.jdk/Contents/Home
Pick the version you want to be the default (i.e 1.6.0_65-b14-462) then:
export JAVA_HOME=/usr/libexec/java_home -v 1.8**After you've successfully install Java8, install dptron with the following command:
pip install mltronsAutoDataPrepIt's important that you replace all the paths that include the folder "Program Files" or "Program Files (x86)" to avoid future problems while running Spark.
If you have Java already installed, you still need to fix the JAVA_HOME and PATH variables. To do that, you need to:
1. Rename "Program Files" with "Progra~1"
2. Rename "Program Files (x86)" with "Progra~2"
Example: "C:\Program FIles\Java\jdk1.8.0_161" --> "C:\Progra~1\Java\jdk1.8.0_161"
After renaming, make sure you have Java 8 installed and the environment variables correctly defined1:
3. Download Java JDK 8 from [Java's official website] (https://www.oracle.com/technetwork/java/javase/downloads/jdk8-downloads-2133151.html)
After installing Java SDK 8, set the following environment variables:
4. JAVA_HOME = C:\Progra~1\Java\jdk1.8.0_161
5. PATH += C:\Progra~1\Java\jdk1.8.0_161\bin
After you've successfully installed and configured Java8, install dptron with the following command:
pip install mltronsAutoDataPrep-
address path of the file
-
local location of the file exist (local pc or s3 bucket)
-
file_format format of the file (csv,excel,parquet)
-
s3 s3 bucket credentials (applicable only if data on s3 bucket)
from mltronsAutoDataPrep.lib.v2.Operations.readfile import ReadFile as rf
res = rf.read(address="test.csv", local="yes", file_format="csv", s3={})-
provide dataframe with threshold of null values
-
return the list of columns containing null values more then the threshold
from mltronsAutoDataPrep.lib.v2.Middlewares.drop_col_with_null_val import DropNullValueCol
res = rf.read("test.csv", file_format='csv')
drop_col = DropNullValueCol()
columns_to_drop = drop_col.delete_var_with_null_more_than(res, threshold=30)
df = res.drop(*columns_to_drop)-
provide dataframe
-
return the list of columns containing same values
from mltronsAutoDataPrep.lib.v2.Middlewares.drop_col_with_same_val import DropSameValueColumn
drop_same_val_col = DropSameValueColumn()
columns_to_drop = drop_same_val_col.delete_same_val_com(res)
df = res.drop(*columns_to_drop)-
Automatically detects features containing Urls
-
Pipeline structure to clean the urls using NLP techniques
from mltronsAutoDataPrep.lib.v2.Pipelines.etl_pipeline import EtlPipeline
etl_pipeline = EtlPipeline()
etl_pipeline.custom_url_transformer(res)
res = etl_pipeline.transform(res)-
Automatically detects features containing date/time
-
Split date time into usefull multiple feautures (day,month,year etc)
from mltronsAutoDataPrep.lib.v2.Pipelines.etl_pipeline import EtlPipeline
etl_pipeline = EtlPipeline()
etl_pipeline.custom_date_transformer(res)
res = etl_pipeline.transform(res)- Using Deep Learning techniques Missing values are filled
from mltronsAutoDataPrep.lib.v2.Pipelines.etl_pipeline import EtlPipeline
etl_pipeline = EtlPipeline()
etl_pipeline.custom_filling_missing_val(res)
res = etl_pipeline.transform(res)-
Automatically detects which column contains skewness
-
Minimize skewness using statistical methods
from mltronsAutoDataPrep.lib.v2.Pipelines.etl_pipeline import EtlPipeline
etl_pipeline = EtlPipeline()
etl_pipeline.custom_skewness_transformer(res)
res = etl_pipeline.transform(res)- Provide list of features in which contains spelling mistakes
from mltronsAutoDataPrep.lib.v2.Pipelines.etl_pipeline import EtlPipeline
etl_pipeline = EtlPipeline()
etl_pipeline.custom_spell_transformer(res,['col1','col2'])
res2 = etl_pipeline.transform(res)
