choose ML alg.txt

see how dep feature w.r.t dep linear or nonlinear if mixed alot then use knn

see data visualiation

usecase:supervisied
        (regression or classification)
for classification use cm,accuracy score,tpr,precision,f1 score to cal
forr reg check acc by Rsqu & adjusted Rsqu

classification:1.log reg (less time to create)
               2.DT (have multiple if else loop)
               3.RF
               4.xgboosting
               5.knn(use when all points mixed completively)
               6.svm
visualisation of data(seaborn(pairplot),matplot) to select ML model 
on graph check each alg can do or not  linear seprarable or not 


for linear seperable then use  log reg(stralight line),  separable (NO OVER LAP then use)
for non linear(have data overlapping can do separable)DT,RF,knn,NN
if mixture alot alot alot   then use knn  first
DT,RF,xgboost,knn req much time TO TRAIN 
when have very very overlap then use KNN(find similarity by using nuclear distance) 
DT,RF,xgboost(use DT),knn(use eli distance)

linear,log assume follow  gaussian  distrbution  if not transform dep feature to gaussian  distrbution get good score