100 days ML

 1. CSV files         i. reading csv files from local storage                df=pd.read_csv('healthcare-dataset-stroke-data.csv')        ii. making your own column instead of pandas default column               df=pd.read_csv('healthcare-dataset-stroke-data.csv',index_col='id')       iii. reading file from some link            import requests            from io import StringIO  url = "link name" headers = {"User-Agent": "Mozilla/5.0 (Macintosh; Intel Mac OS X 10.14; rv:66.0) Gecko/20100101 Firefox/66.0"} req = requests.get(url, headers=headers) data = StringIO(req.text) pd.read_csv(data)    iv.if there is no column name or column name become as a temple becoming first row value      pd.read_csv('test.csv',header=1) v.if you want to select only some of the colum...

 1Data Engineer:       task:           1.collect all the data required to make machine learning algorithm,          2. make data warehouse from database:                      if you want to make change the existing database you need to create datawarehouse and modify the existing data for the future use it is done by data engineer    3. make pipeline for the use of that data while deploying to the server 2. Data Analyst:       task:          1.data analysis and processing          2. presentation about data to CEO and other member 3. data scientist:     task:         1.full stack data analyst and machine learning algorithm 4.ML Engineer:     task:         1.deploying the built model to server         2. mai...

 instance model:    the point to be predicted or to be analyzed checks its neighbor point and compares its probability or output value: If the output value is success or positive in case of compared value than result of that point will be also success else fails model based:  it creates certain pattern and if the predicted value falls under the desired pattern model gives success result else not success

Batch learning:   you create your model in the developmental site and you deploy that model to online server . It worked but only with new data only  take an example of Netflix recommendation system: there is update of movies and your model is working with old data and already new movies are added so theremust be provision to update your batch system model.  If you update your model in 24-24 hours by collecting new data from particular deployment system or in 1 week than also there are many problems of this system such as: 1. lots of data     there could be lots of data being piled up in your model and there is possibility at some period of time model cant process that much amount of data 2. hardware limitation:    if you are in online service and one person is in Mustang no internet service and he stays there for 1 week and is currently watching Netflix let us assume : your model has already updated in 24 hours and muktinath person still is in old mod...