10. Python - Data Analytics - Scikit-Learn

MODEL VALIDATION:

Splitting Data with Scikit_Learn:

from sklearn.model_selection import train_test_split

train_X, val_X, train_y, val_y = train_test_split(X, y, random_state = 0)

// splits data into training and validation data. must define X and y beforehand, which is the features and the prediction.

melbourne_model.fit(train_X, train_y)

# get predicted prices on validation data

val_predictions = melbourne_model.predict(val_X)

print(mean_absolute_error(val_y, val_predictions))


Calculating MAE (Mean Absolute Error):

from sklearn.metrics import mean_absolute_error

predicted_home_prices = melbourne_model.predict(X)

mean_absolute_error(y, predicted_home_prices)

// MAE is the sum average of all absolute difference between the actual and the prediction value.


Overfitting and Underfitting:

// Overfitting is too many leaves in decision tree. Might seem like it gets accurate prediction but it becomes far off when using new data to predict.

// Underfitting is too few leaves in decision tree. This makes decision tree too broad and doesn't capture many distinctions. Not accurate.

// To find best balance. Find the sweet spot in number of leaves to use that generate lowest MAE.


Create a function to use different types of leaf nodes:

from sklearn.tree import DecisionTreeRegressor


def get_mae(max_leaf_nodes, train_X, val_X, train_y, val_y):

    model = DecisionTreeRegressor(max_leaf_nodes=max_leaf_nodes, random_state=0)

    model.fit(train_X, train_y)

    preds_val = model.predict(val_X)

    mae = mean_absolute_error(val_y, preds_val)

    return(mae)


Print and compare leave node options and their MAE results:

# compare MAE with differing values of max_leaf_nodes

for max_leaf_nodes in [5, 50, 500, 5000]:

    my_mae = get_mae(max_leaf_nodes, train_X, val_X, train_y, val_y)

    print("Max leaf nodes: %d  \t\t Mean Absolute Error:  %d" %(max_leaf_nodes, my_mae))


// after knowing the best leaf node amount. We can refit a new model using all the original X and y data and not use train_X or train_y

Comments

Post a Comment

Popular posts from this blog

2. FreeCodeCamp - Dynamic Programming - Learn to Solve Algorithmic Problems & Coding Challenges

Dynamic Programming - Parts      PART 1 - Memoization     P ART 2 - Tabulation Dynamic Programming with Fibonacci Example Recursion: Slow because time complexity is O(2^n) func fibR (x int ) int {     if x <= 2 {         return 1     }     return fibR (x- 1 ) + fibR (x- 2 ) } Dynamic Way Example Recursion with Stored Value:  // Fast because it removes duplicate processes by checking map. // Tail recursion-ish // Results are hitchhiking each recursive call func fibR2 (x int , y map [ int ] int ) int {     if _ , ok := y[x]; ok {         return y[x]     }     if x <= 2 {         return 1     }     y[x] = fibR2 (x- 1 , y) + fibR2 (x- 2 , y)     return y[x] }
Read more

20. Data Analytics - Analyze Data to Answer Questions - Week 1

Goal of analysis is to identify trends and relationships within data so you can accurately answer the questions you're asking. Most data is organized in tables, always have data in the right format. Definition : Analysis // process used to make sense of the data collected Sorting // involves arranging data into a meaningful order to make it easier to understand, analyze, visualize Outliers // data points that are very different from similarly collected data and might not be reliable values Filtering // extracting specific data that meet certain criteria while hiding the rest Customized sort order // when you sort data in a spreadsheet using multiple conditions The 4 phases of analysis:          1. Organize data // getting the right data in the right size and order                     - Sorting : arrange data in meaningful order                     - Filtering : can help find errors or outliers          2. Format and adjust data // filtering or sorting or compiling data in a better
Read more

3. Algorithms - Selection Sort

Selection sort sorts an array by finding the lowest value element and appending it to an empty array. The algorithm will shrink the unsorted array and repeat the process recursively until all unsorted elements are used up. package main func selectionSort (list *[] int ) {     sorted := make ([] int , 0 , len (*list))     unSorted := *list     for i := 0 ; i < len (*list); i++ {         min := unSorted[ 0 ]         index := 0         for i , e := range unSorted {             if e <= min {                 min = e                 index = i             }         }         sorted = append (sorted, (unSorted)[index])         if index == 0 {             unSorted = unSorted[ 1 :]         } else if index == ( len (*list) - 1 ) {             unSorted = unSorted[:index]         } else {             unSorted = append (unSorted[ 0 :index], unSorted[index+ 1 :]...)         }     }     * list = sorted } /*func main() {     x := []int{4, 6, 3, 2, 9, 7, 3, 5}     fmt.Print
Read more