Page 133 - Artificial Intellegence_v2.0_Class_12
P. 133
p r ed iction= mod elT oI mp lement.p r ed ict( [ [ 5 .0 , 2 .0 , 1 , 0 .2 ] ] ) # p r ed iction w ith new v alu es
(
p r int p r ed iction)
Output
[ ' I r is- v er sicolor ' ]
[ ' I r is- setosa' ]
ode can be accessed through the link https colab.research.google.com
drive p m p g m d o m tqa usp sharing
Metrics of Model Quality—Simple Maths
no ing the quality of predictions is important in assessing the performance of a machine learning model. he
performance metrics, derived through simple mathematical calculations hich provides ob ective measures of ho ell
the predictions align ith the e pected outcomes. etrics like classification accuracy and root mean squared error allo
for comparisons bet een different models, data transformations, and configurations. hese metrics play a vital role in
implementing machine learning algorithms and are based on minimi ing or ma imi ing an ob ective function kno n as
a loss function.
A loss function is used by machines to learn. It's a ay of determining ho ell a certain algorithm models the data.
If the forecasts are too far off from the actual findings, the loss function ill return a very large number. oss function
learns to lo er prediction error over time ith the help of some optimisation ob ective function.
In machine learning, there is no such thing as a one si e fits all loss function. he type of machine learning method
used, the ease of calculating derivatives, and to some e tent, the number of outliers in the dataset, all play a role in
selecting a loss function for a certain task.
Depending on the type of learning ob e're dealing ith, loss functions can be divided into t o categories regression
losses and classification losses. In classification, the output is predicted from a set of finite categorical values. or e ample,
given a large data set of photographs of hand ritten numbers, categorising them into one of digits. egression,
on the other hand, is concerned ith predicting a continuous value, such as given the oor area, number of rooms and
room si e, predicting the price of the house.
Classi cation Regression
og oss ean quare rror uadratic oss
ocal oss ean Absolute rror
uber oss mooth ean
divergence elative ntropy
Absolute rror
ponential oss og ash oss
inge oss uantile oss
MS E ( Mean S q uar ed Er r or )
he ean quared rror ) is the most basic and idely used loss function, and it is frequently taught in achine
earning courses. alculate the difference bet een the model's predictions and actual values, square it, and average it
across the entire dataset to get the value of . is given by the equation
C apstone P roj e ct 131

