Review of Python Courses (Part 13)

In Part 12, I summarized my Datacamp courses 35-37. Today I will continue with the next three.

As a reminder, I introduced you to my recent work learning Python here.

My course #38 was Web Scraping in Python. This gets complicated with some objected-oriented stuff that still throws me for a loop (no pun intended). I don’t think I will be using this anytime soon so I skimmed it in this review:

• Web scraping with Python
• HyperText Markup Language (HTML)
• HTML tags and attributes
• Crash course X
• Off the beaten XPath
• Introduction to the scrapy Selector (from scrapy import Selector)
• “Inspecting the HTML”
• CSS locators
• Attribute and text selection
• Getting ready to crawl
• Scraping for reals
• A classy spider (from scrapy.crawler import CrawlerProcess)
• A request for service
• Move your bloomin’ parse
• Capstone

My course #39 was Working with the Class System in Python. Like #38, this gets thick. The course covers:

• Intro to Object Oriented Programming (OOP) in Python
• Introduction to NumPy internals
• Introduction to objects and classes
• Deep dive on classes
• __Init__ializing a class
• Methods in classes
• Working with a dataset to create dataframes
• Renaming columns and the five-figure summary
• OOP best practices
• Inheritance: is-a versus has-a
• Inheritance with DataShells
• Composition
• Wrapping up OOP

My course #40 was Sentiment Analysis in Python. This course covers:

• What is sentiment analysis?
• Sentiment analysis types and approaches (from textblob import TextBlob)
• Let’s build a word cloud (from wordcloud import WordCloud)!
• Bag-of-words (from sklearn.feature_extraction.text import CountVectorizer)
• Getting granular with n-grams
• Build new features from text (from nltk import word_tokenize)
• Can you guess the language (from langdetect import detect_langs)?
• Stop words (from sklearn.feature_extraction.text import ENGLISH_STOP_WORDS)
• Capturing a token pattern [.isalpha(), .isdigit(), .isalnum()]
• Stemming and lemmatization (from nltk.stem import PorterStemmer, WordNetLemmatizer)
• TfIdf: more ways to transform text (from sklearn.feature_extraction.text import TfidfVectorizer)
• Let’s predict the sentiment (from sklearn.linear_model import LogisticRegression)!
• Did we really predict the sentiment well (from sklearn.metrics import accuracy_score, confusion_matrix)?
• Logistic regression: revisited
• Bringing it all together

I will review more classes next time.

Review of Python Courses (Part 12)

In Part 11, I summarized my Datacamp courses 31-34. Today I will continue with the next three.

As a reminder, I introduced you to my recent work learning Python here.

My course #35 was Intermediate Data Visualization with Seaborn. This course covers:

• Introduction to Seaborn [histogram vs. sns.distplot()]
• Using the distribution plot
• Regression plots in Seaborn [sns.regplot(), sns.lmplot()]
• Using Seaborn styles [sns.set_style(), sns.despine()]
• Colors in Seaborn
• Customizing with matplotlib (using Axes)
• Categorical plot types
• Regression plots [sns.regplot(), sns.residplot()]
• Matrix plots [sns.heatmap(pd.crosstab())]
• Using FacetGrid, factorplot, lmplot
• Using PairGrid and pairplot
• Using JointGrid and jointplot
• Selecting Seaborn plots

My course #36 was Introduction to Data Visualization with Seaborn (taking #35 before this was an oversight on my part, but everything ended up okay). This course covers:

• Introduction to Seaborn
• Using pandas with Seaborn
• Adding a third variable with hue
• Introduction to relational plots and subplots
• Customizing scatter plots
• Introduction to line plots
• Count plots and bar plots [sns.catplot()]
• Creating a box plot
• Point plots
• Changing plot style and color
• Adding titles and labels (FacetGrid vs. AxesSubplot)

My course #37 was Unsupervised Learning in Python. This course covers:

• Unsupervised learning (from sklearn.cluster import KMeans)
• Evaluating a clustering
• Transforming features for better clustering (from sklearn.preprocessing import StandardScaler)
• Visualizing hierarchies (from scipy.cluster.hierarchy import linkage, dendrogram)
• Cluster labels in hierarchical clustering
• t-SNE for 2-dimensional maps (from sklearn.manifold import TSNE)
• Visualizing the PCA transformation (from sklearn.decomposition import PCA)
• Intrinsic dimension
• Dimension reduction with PCA (from sklearn.decomposition import TruncatedSVD)
• Non-negative matrix factorization (NMF) (from sklearn.decomposition import NMF)
• NMF learns interpretable parts
• Building recommender systems using NMF (From sklearn.preprocessing import normalize)

I will review more classes next time.

Review of Python Courses (Part 11)

In Part 10, I summarized my Datacamp courses 28-30. Today I will continue with the next four.

As a reminder, I introduced you to my recent work learning Python here.

My course #31 was Customer Analytics and A/B Testing in Python. This course covers:

• What is A/B testing?
• Identifying and understanding KPIs
• Exploratory analysis of KPIs
• Calculating KPIs—a practical example
• Working with time series data in pandas
• Creating time series graphs with matplotlib
• Understanding and visualizing trends in customer data
• Events and releases
• Introduction to A/B testing
• Initial A/B test design
• Preparing to run an A/B test
• Calculating sample size
• Analyzing the A/B test results
• Understanding statistical significance (get_pvalue, get_ci)
• Interpreting your test results

My course #32 was Machine Learning with Tree-Based Models in Python. This course covers:

• Decision-tree for classification (from sklearn.tree import DecisionTreeClassifier)
• Classification-tree learning
• Decision-tree for regression
• Generalization error (bias-variance tradeoff)
• Diagnosing bias and variance problems
• Ensemble learning
• Bagging (from sklearn.ensemble import BaggingClassifier)
• Out of bag evaluation
• Random forests
• AdaBoost (from sklearn.ensemble import AdaBoostClassifier)
• Gradient boosting (from sklearn.ensemble import GradientBoostingRegressor)
• Stochastic gradient boosting
• Tuning a CART’s hyperparameters
• Tuning an RF’s hyperparameters

My course #33 was Introduction to PySpark. This is a data engineering course—a field in which I found myself not very enthusiastic. This course covers:

• What is Spark, anyway?
• Using Spark in Python
• Using dataframes
• Joining
• Maching learning pipelines
• Data types
• Strings and factors

My course #34 was Cleaning Data with PySpark. This course covers:

• Intro to data cleaning with Apache Spark
• Immutability and lazy processing
• Understanding Parquet
• Dataframe column operations
• Conditional dataframe column operations
• User defined functions
• Partitioning and lazy processing
• Caching
• Improve import performance
• Cluster sizing tips
• Performance improvements
• Introduction to data pipelines
• Data handling techniques
• Data validation
• Final analysis and delivery

I will review more classes next time.

Review of Python Courses (Part 10)

In Part 9, I summarized my Datacamp courses 25-27. Today I will continue with the next three.

As a reminder, I introduced you to my recent work learning Python here.

My course #28 was Supervised Learning with scikit-learn. This course covers:

• Supervised learning
• Exploratory data analysis [pd.plotting.scatter_matrix()]
• The classification challenge (creating arrays, from sklearn.neighbors import KNeighborsClassifier)
• Measuring model performance (from sklearn.model_selection import train_test_split, datasets)
• Introduction to regression (from sklearn.linear_model import LinearRegression)
• The basics of linear regression
• Cross-validation (from sklearn.model_selection import cross_val_score)
• Correlation
• Simple regression (from scipy.stats import linregress) and its limits
• Regularized regression (from sklearn.linear_model import Ridge, Lasso)
• How good is your model (from sklearn.metrics import classification_report, confusion_matrix)?
• Logistic regression and the ROC curve (from sklearn.metrics import roc_curve)
• Area under the ROC curve
• Hyperparameter tuning (from sklearn.model_selection import GridSearchCV)
• Hold-out set for final evaluation
• Preprocessing data [pd.get_dummies(df)]
• Handling missing data (from sklearn.preprocessing import Imputer, from sklearn.pipeline import Pipeline)
• Centering and scaling (from sklearn.preprocessing import scale, StandardScaler)

My course #29 was Introduction to Natural Language Processing in Python. This course covers:

• Introduction to regular expressions
• Introduction to tokenization (from nltk.tokenize import word_tokenize, sent_tokenize)
• Advanced tokenization with regex
• Charting word length with nltk
• Word counts with bag-of-words (from collections import Counter)
• Simple text preprocessing (from nltk.corpus import stopwords, from nltk.stem import WordNetLemmatizer)
• Introduction to gensim (from gensim.corpora.dictionary import Dictionary)
• Tf-idf with gensim (from gensim.models.tfidfmodel import TfidfModel)
• Named entity recognition
• Introduction to SpaCy
• Multilingual NER with polyglot (from polyglot.text import Text)
• Classifying fake news using supervised learning with NLP
• Building word count vectors (from sklearn.feature_extraction.text import CountVectorizer, TfidfVectorizer)
• Training and testing a classification model with scikit-learn (from sklearn.naive_bayes import MultinomialNB)
• Simple NLP, complex problems

My course #30 was Building Chatbots in Python. This course covers:

• Introduction to conversational software (respond function, sleep method from time module)
• Creating a personality
• Text processing with regular expressions
• Understanding intents and entities (re.compile)
• Word vectors
• Intents and classification (from sklearn.svm import SVC)
• Entity extraction
• Robust NLU with Rasa (from rasa_nlu.converters import load_data)
• Virtual assistants and accessing data
• Exploring a DB with natural language
• Incremental slot filling and negation
• Stateful bots
• Asking questions and queuing answers
• Frontiers of dialog technology

I will review more classes next time.

Review of Python Courses (Part 9)

In Part 8, I summarized my Datacamp courses 22-24. Today I will continue with the next three.

As a reminder, I introduced you to my recent work learning Python here.

My course #25 was Exploratory Data Analysis in Python (Part 2). This course covers:

• Dataframes and series
• Clean and validate (inplace arg)
• Filter and visualize
• Probability mass functions
• Cumulative distribution functions (probability < x)
• Comparing and modeling distributions
• Exploring (scatter plot: transparency, market size, jittering, zoom) and visualizing relationships (violin, box plot)
• Correlation
• Simple regression (from scipy.stats import linregress) and its limits
• Multiple regression
• Visualizing regression results
• Logistic regression

My course #26 was Regular Expressions in Python. Once into regex, this material gets very complex yet very powerful:

• Introduction to string manipulation
• String operations (selecting portions of a particular word)
• Finding and replacing
• Positional formatting (method to format percentages)
• Formatted string literal (escape sequences)
• Template method (from string import Template)
• Introduction to regular expressions
• Repetitions
• Regex metacharacters
• Greedy vs. non-greedy matching
• Alternation and non-capturing groups
• Backreferences
• Lookaround

My course #27 was Introduction to Deep Learning in Python. This course covers:

• Introduction to deep learning
• Forward propagation
• Activation functions
• Deeper networks
• The need for optimization
• Gradient descent
• Backpropagation [in practice]
• Creating a Keras model
• Compiling and fitting a model
• Classification models
• Using models
• Understanding model optimization
• Model validation
• Thinking about model capacity
• Stepping up to images

I will review more classes next time.

Review of Python Courses (Part 8)

In Part 7, I summarized my Datacamp courses 19-21. Today I will continue with the next three.

As a reminder, I introduced you to my recent work learning Python here.

My course #22 was Statistical Thinking in Python (Part 2). This course covers:

• Optimal parameters [statistical inference using scipy.stats, statsmodels, or hacker stats with numpy; plt.margins() ]
• Linear regression by least squares [slope, intercept = np.polyfit() ]
• The importance of exploratory data analysis: Anscombe’s quartet (generating and plotting line of best fit)
• Generating bootstrap replicates [ecdf() written in my course (prequel) #14]
• Bootstrap confidence intervals
• Pairs bootstrap
• Formulating and simulating a hypothesis (permutation sample)
• Test statistics and p-values (permutation replicate)
• Bootstrap hypothesis tests
• A/B testing
• Test of correlation

My course #23 was Introduction to Financial Concepts in Python. This course covers:

• Fundamental financial concepts (calculating return on investment and compound interest)
• Present and future value [np.pv(), np.fv() ]
• Net present value and cash flows [np.npv(rate= , values=np.array([]) ) ]
• Common profitability analysis methods [np.npv(), np.irr(np.array([]) ) ]
• Weighted average cost of capital
• Comparing two projects of different life spans (EAA)
• Mortgage basics [np.pmt(rate, nper, pv) ]
• Amortization, principal, and interest (simulating periodic mortgage payments)
• Home ownership, equity, and forecasting (cumulative operations in numpy)
• Budgeting project proposal [constant cumulative growth with np.repeat(), calculating monthly expenses]
• Net worth and valuation in your personal financial life
• The power of time and compound interest

My course #24 was Introduction to Portfolio Risk Management in Python. This course covers:

• Financial returns
• Mean, variance, and normal distributions (scaling volatility)
• Skewness and kurtosis (from scipy.stats import skew, kurtosis, Shapiro-Wilk test)
• Portfolio composition (calculating market-cap weights)
• Correlation and covariance (calculating portfolio volatility)
• Markowitz portfolios (MSR and GMV)
• The capital asset pricing model (calculating Beta)
• Alpha and multi-factor models (Fama-French 3-factor model)
• Expanding the 3-factor model (Fama-French 5-factor model)
• Estimating tail risk (historical drawdown, historical/conditional VaR)
• VaR extensions
• Random walks (Monte Carlo simulations)

Review of Python Courses (Part 7)

In Part 6, I summarized my Datacamp courses 16-18. Today I will continue with the next three.

As a reminder, I introduced you to my recent work learning Python here.

My course #19 was Manipulating DataFrames with pandas. This course covers:

• Indexing DataFrames (using square brackets, using .loc, using .iloc, selecting certain columns with [[ ]] )
• Slicing DataFrames (R boundary included with .loc but not .iloc, slicing with one/two brackets gets Series/df)
• Filtering DataFrames
• Transforming DataFrames [vectorized computations in numpy without loops, .map() for index, .apply() for Series]
• Indexed objects and labeled data (name attribute for index and columns attributes)
• Hierarchical indexing (sorting MultiIndex)
• Pivoting DataFrames [.pivot(index= , columns= , values= ) ]
• Stacking and unstacking DataFrames (pivoting doesn’t work well on MultiIndex so unstack to move index to column)
• Melting DataFrames [reverses .pivot() ]
• Pivot tables
• Categoricals and groupby
• Groupby and aggregation/transformation
• Iterating over and filtering groupby object
• Understanding the column labels
• .idxmax() and .idxmin() (row/column label where max/min value located)
• .T attribute (transposes numpy array)
• Reshaping DataFrames for visualization
• Making a histogram (bins, range, normalizing)

My twentieth course was Manipulating Time Series Data in Python. This course has lots of good information for backtesting:

• How to use dates and times with pandas ( [sequences of] timestamp and period objects)
• Indexing and resampling time series [selecting missing ‘price’ values, .asfreq() ]
• Lags, changes, and returns for stock price series [.shift(), n-period % chg, .diff(), .pct_change(), stock price chg in df]
• Compare time series growth rates ( .iloc as abs ref, normalizing series, concat prices and .dropna, perf vs. benchmark)
• Changing the time series frequency: resampling
• Upsampling and interpolation with .resample()
• Downsampling and aggregation (plotting resample data with ax)
• Rolling window functions with pandas (plotting price and moving average, plotting multiple rolling metrics)
• Expanding window functions with pandas (calculating running return, running rate of return)
• Relationships between time series: correlation
• Select index components and import data
• Build a market-cap weighted index
• Evaluate index performance
• Index correlation and exporting to Excel

My course #21 was Working with Dates and Times in Python. This course covers:

• Dates in Python
• Math with dates (time delta)
• Turning dates into strings
• Adding time to the mix
• Printing and parsing datetimes (no time printed from datetime object)
• Working with durations
• UTC offsets
• Time zone database (from dateutil import tz)
• Starting Daylight Saving Time
• Ending Daylight Saving Time [ .datetime_ambiguous() and .enfold() for ambiguous times]
• Reading date and time data in Pandas (loading datetimes with parse_dates [or manually with .to_datetime() ])
• Summarizing datetime data in Pandas (alternative to for loop)
• Additional datetime methods in Pandas
• Index correlation and exporting to Excel

Review of Python Courses (Part 6)

In Part 5, I summarized my Datacamp courses 13-15. Today I will continue with the next three.

As a reminder, I introduced you to my recent work learning Python here.

My course #16 was Introduction to Data Science in Python. This course covers:

• Creating variables
• What is a function?
• What is pandas?
• Selecting columns
• Select rows with logic
• Creating line plots
• Adding labels and legends
• Adding some style (line color, width, style, markers, template)
• Making a scatter plot (marker transparency)
• Making a bar chart (horizontal, error bars, stacked)
• Making a histogram (bins, range, normalizing)

My course #17 was Joining Data with Pandas. This course covers:

• Inner join (changing df values with .loc accessor)
• One to many relationships
• Merging multiple DataFrames
• Left join (count number of rows in a column with missing data)
• Right and outer joins
• Merging a table to itself (i.e. self join)
• Merging on indexes
• Filtering joins (semi-joins, anti-joins)
• Concatenate DataFrames together vertically [.append()]
• verify_integrity=True identifies accidental duplicates while validate arg helps to identify relationship type
• Using merge_ordered() (for ordered/time-series data and to fill in missing values)
• Using .merge_asof() (matches on nearest-value rather than equal-value columns)
• Selecting data with .query()
• Reshaping data with .melt()

Introduction to Linear Modeling in Python was my eighteenth course. This covers:

• Introductory concepts about models (interpolation, extrapolation)
• Visualizing linear relationships [object-oriented (OOP) approach to matplotlib]
• Quantifying linear relationships (covariance, correlation, normalization)
• What makes a model linear (Taylor series, overfitting, defining function to plot graph)
• Interpreting slope and intercept
• Model optimization (RSS: sum of squared residuals)
• Least-squares optimization (by numpy, Scipy, Statsmodels)
• Modeling real data
• The limits of prediction
• Goodness of fit (deviations, residuals, and R-squared in code)
• Standard error (RMSE measures spread of residuals whereas SE measures uncertainty in model params)
• Inferential statistics concepts
• Model estimation and likelihood
• Model uncertainty and sample distributions (bootstrap in code)
• Model errors and randomness