Jan-10-2020, 12:13 PM
(This post was last modified: Jan-10-2020, 12:14 PM by jefsummers.)
Flask is good for developing a web interface for your application. There are a variety of GUI interfaces available as well. Tkinter is very common, I like wxPython.
That out of the way, you will want to use Pandas as it is foundational to a lot of the other packages. Get very comfortable with the notation, slicing, and manipulating of dataframes.
You can do a lot with SciKit Learn and that is where I would start for data analysis. Get comfortable with the Linear Regression, Polynomial regression, K Means, K Nearest Neighbor, and other techniques there.
Moving to machine learning, I'd use TensorFlow, though there are others. Keras plays nicely with TensorFlow and extends your capabilities.
A little depends on where you plan to do your data analysis. Reasonable sized datasets on your own system can be done with just about any of the available packages. If you have a massive dataset (all Medicare claims for 2018, all taxi rides in NYC 2010-2012, etc.) you will likely use a cloud system with virtual machines. Correct me if wrong, but IBM/Watson uses PyTorch more than TensorFlow. Google Cloud Services uses TensorFlow.
Please ask more specifics if you need, will do my best to answer or direct you.
J
Oh, and Pandas will webscrape for the right formats - will pull all tables on a webpage into an array of dataframes. For more sophisticated webscraping Beautiful Soup seems to be the most popular right now.
That out of the way, you will want to use Pandas as it is foundational to a lot of the other packages. Get very comfortable with the notation, slicing, and manipulating of dataframes.
You can do a lot with SciKit Learn and that is where I would start for data analysis. Get comfortable with the Linear Regression, Polynomial regression, K Means, K Nearest Neighbor, and other techniques there.
Moving to machine learning, I'd use TensorFlow, though there are others. Keras plays nicely with TensorFlow and extends your capabilities.
A little depends on where you plan to do your data analysis. Reasonable sized datasets on your own system can be done with just about any of the available packages. If you have a massive dataset (all Medicare claims for 2018, all taxi rides in NYC 2010-2012, etc.) you will likely use a cloud system with virtual machines. Correct me if wrong, but IBM/Watson uses PyTorch more than TensorFlow. Google Cloud Services uses TensorFlow.
Please ask more specifics if you need, will do my best to answer or direct you.
J
Oh, and Pandas will webscrape for the right formats - will pull all tables on a webpage into an array of dataframes. For more sophisticated webscraping Beautiful Soup seems to be the most popular right now.