Skip to content
/ ML Public

Machine Learning Experiments with scikit-learn, Deep learning with Keras, TensorFlow and Pytorch. Data Science examples for various datasets and competitions from Kaggle and Analytics Vidhya.

License

Notifications You must be signed in to change notification settings

greed2411/ML

Repository files navigation

Sometimes Deep Sometimes Learning

made-with-python Maintenance Ask Me Anything ! License GitHub issues Awesome Badges

Experimenting with Machine Learning and Deep Learning in Python.

I mostly use Jupyter Notebook, with Anaconda distribution of Python when it comes to DS/ML/DL.

Kaggle Profile : jaivarsan.

Analytics Vidhya Profile : greed2411.

Libraries used :

  • scikit-learn : for purely Machine Learning Algorithms and preprocessing.
  • pandas : for data analysis purposes.
  • seaborn : for beautiful visualizations.
  • plotly : for time series graphs.
  • gensim : for NLP.
  • keras : for deep learning purposes.
  • tensorflow : for keras backend.
  • pytorch : cause the hype is real.

Papers Implemented :

K. Altun, B. Barshan, and O. Tunçel, Comparative study on classifying human activities with miniature inertial and magnetic sensors, Pattern Recognition, 43(10):3605-3620, October 2010.

for UCI ML repositories : Daily and Sports Activities Data Set report and files available over here

Time for visualizations:

On Miscellaneous

Indian map plot of all the 498 cities mentioned in the NDL dataset from Kaggle:

india

Heat map from california house pricing dataset:

california

Object Detection using TensorFlow for both pictures and Video, for our Microprocessor Project using RPi:

Me and my friends, during a trek:

Bagayam

Friend of mine got caught on the RPi cam feed:

objvid

The famous iris dataset

  • Andrews curves

andrewcurves

  • Parallel Plotting pp

  • Pairplot Pariplot

Data is Beautiful, Not Saturn's rings

saturn

On Healthcare

Breast Cancer

countplot

histogram

Melanoma Prediction

melanoma

Cycling and Calories dataset

cyclingregression

On Entertainment

The time when i solved math memes from facebook

mathmeme

In [28]: best_model_pred

Out[28]:

['7+2+5 = 107687\n',
 '7+2+5 = 191991\n',
 '7+2+5 = 202541\n',
 '7+2+5 = 18908\n',
 '7+2+5 = 183652\n']

On Business

Scatterplot on Relationship between Sales because of TV, Radio and Newspaper.

scatterplot

Plotly graphs of the recent trends in Bitcoin world, intereactive time series graphs.

bitcoin trend

btcexchangemarts

Just to give you a tease:

This is the first dataset I handled, xD.

While working on NDL, A Kaggle dataset

I made the following observations:

Analysis made on the data set as of June 2017

  • Number of cities : 493

  • Number of states : 29

  • Most number of cities are in the state: UTTAR PRADESH

  • Number of cities in UTTAR PRADESH : 63

  • Least number of cities belong to these states and their counts

    HIMACHAL PRADESH             1
    CHANDIGARH                   1
    TRIPURA                      1
    MIZORAM                      1
    NAGALAND                     1
    MANIPUR                      1
    MEGHALAYA                    1
    ANDAMAN & NICOBAR ISLANDS    1
    
  • There are two Aurangabad(s) in the nation,

    • one belonging to BIHAR
    • second one belonging to MAHARASHTRA
  • Top 5 States with the maximum number of Cities

      UTTAR PRADESH     63
      WEST BENGAL       61
      MAHARASHTRA       43
      ANDHRA PRADESH    42
      TAMIL NADU        32
    
  • States vs City counts

    States&citycounts

  • States vs District counts

    States&districtcounts

  • Each city and it's district number plot

    City&districtnumber

    This graph made me analyse and conclude that

    • The most common district numbers are 11, 9 and 12 not the conventional 1, 2 and 3.

    • District Number, and their occurences and percentage they contribute to the total district count. Example : District Number 11, there are 37 districts in our India numbered 11, which contributes to 7.51% of total number of districts in India.

                                   District Counts  Percentage Index
       District Number                                   
       11                            37              7.51
       9                             26              5.27
       12                            24              4.87
       1                             22              4.46
       3                             22              4.46
       21                            21              4.26
      
    • 95.94% of districts have their district number value which is less than 50

  • District numbers and their frequency

    districtcounts

    • The above graphs tells us that there are no district numbers from 72 to 98 and few numbers here and there in the 40s & 50s

      Actual missing district numbers :

       40, 42, 43, 45, 51, 53, 55, 56, 58, 67, 69, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98
      
  • Missing State Codes

     11, 12, 25, 26, 30, 31
    

About

Machine Learning Experiments with scikit-learn, Deep learning with Keras, TensorFlow and Pytorch. Data Science examples for various datasets and competitions from Kaggle and Analytics Vidhya.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published