Data science, machine learning, and analytics without coding

Use exploratory data analysis to gain data understanding, check data quantity and labeling, and identify quality issues that can derail modeling.

Take a KNIME tour and learn to import data with the file reader, accessing csv and excel files within a workflow, and preview the dataset in the file table.

Explore the five core data types—double (and float), integer, string, date/time, and boolean—in KNIME and any data analysis tool, with notes on math, comparisons, and joins.

Explore KNIME's group by node to aggregate data by date and shift, using sums and counts to reveal insights, sort groups, and compare performance across periods.

Pivot data by day and shift to compute failure rate with a simple math formula. Create pivot tables to compare first, second, and third shifts for actionable insights.

Explore summary statistics from the statistics node, including mean, standard deviation, variance, skew, and a histogram to reveal data distribution across three shifts and guide data cleaning.

Plot data in KNIME with dates on the x axis and the reject count on the y axis, using interactive filters. Identify items to clean before modeling.

Learn data cleaning through exploration, iterative fixes of strings and duplicates, and handling missing data and outliers to prepare data for modeling.

Convert a date-like string to a date type using the string to date step in time series. Use the format d-MM-yy, noting capital M denotes month and the dash matters.

Learn practical string manipulation to clean data: convert numbers to strings, remove periods and commas, strip exponential notation, and pad identifiers to a fixed length.

Learn to clean up your data workflow by combining nodes into metanodes, label steps, and fit large programs into a single node you can open to inspect.

Explore string manipulation to unpack a 14-character ident into part number and color, group items, and count occurrences, then validate substrings to reveal structure.

Use KNIME's rule engine to map color codes to names with if-else statements and create a color text column for colors like orange, blue, red, and black.

Apply row filtering to remove unwanted data by color-based and rule-based criteria, using pattern matching and membership checks to exclude red and black entries and keep orange and blue.

Discover how KNIME's missing value node replaces missing data, with options such as fixed value, average, previous, and moving average for strings and doubles.

Modeling introduction emphasizes building blocks and data readiness, using the crisp dm framework to define the business problem and select appropriate models such as linear regression and random forest.

Cycle through the cross-industry standard process for data mining (crisp-dm) to solve data problems with an iterative flow of business and data understanding, data preparation, modeling, evaluation, and deployment.

Explore what machine learning is, including supervised, unsupervised, and reinforcement learning, and distinguish classification versus regression with examples like predicting categories and house prices.

Develop a machine learning model to predict win outcomes for the Jacksonville Sales Agency using twelve months of Salesforce export data, focusing on opportunities, stage, probability, amount, and office indicators.

Explore how the random forest builds multiple decision trees from bootstrap subsamples and voting, delivering robust, accurate predictions with interpretable insights in KNIME.

Explore how to concatenate and union multiple monthly data sets into a single complete data set, aligning months from january to december and exporting a clean file with headers.

Joiner shows how to build a dataset with outcomes for modeling by using a left join on opportunity ID, creating a one indicator for won or lost, and filtering proposals.

Filter out open opportunities and clean dollar amounts to prepare data for binning. Use five bins to group numeric features like amount and days open, comparing equal-width and equal-frequency approaches.

Partition data into 80% training and 20% testing. Build a random forest with a learner and predictor, select the target column and features, and score with out-of-bag predictions.

Analyze model scoring using accuracy, precision, recall, and confusion matrix to understand false positives and false negatives and the business usefulness of the model.

Explore how outsourced collections call centers maximize cash per hour by using a crisp d.m. process and evaluating contact rate, lists, and dialing methods.

Explore how ordinary least squares builds linear regression with one or more independent variables, using coefficients to predict cash per hour, and verify assumptions like linearity, normality, homoscedasticity, and independence.

Learn to implement a linear regression model, diagnose issues using a correlation matrix, and select predictors by removing highly correlated features to predict cash, with 80/20 partition and R-squared.

Apply backward stepwise regression to refine a linear regression model by removing non-significant predictors based on p-values, monitor r-squared, and guard against overfitting.

Evaluate all four linear regression assumptions, confirm linear relationships with scatterplots, compute and standardize residuals (z-scores), and assess heteroscedasticity and normality with residual plots and histograms.

Evaluate linear regression by interpreting r-squared and adjusted r-squared to gauge variance explained, using the rule-of-thumb thresholds of about 0.1, 0.5, and 0.75–0.8.

Explore how to turn predictions into business impact by prioritizing close rate and contact rate, and evaluating training hours and account scores for real value.

Apply data understanding and business understanding to a K means clustering example. Analyze demographic data, repeat customers, and customer acquisition cost in a cross industry standard data mining process.

Demonstrate k-means clustering as an unsupervised method that minimizes within-cluster distance using Euclidean distance, updates centroids, and discusses elbow and scree plots for choosing k.

Perform a first k-means clustering on continuous data, clean the dataset by filtering out extra columns and missing rows, and examine five clusters by income, credit score, and market score.

Normalize data with a z-score before applying k-means to ensure Euclidean distance reflects the income axis, highlighting why scaling incomes and credit scores matters for clustering.

Assess clusters by analyzing income, credit score, market score, and repeat rate, then group by cluster to compute repeat percent and compare sizes for targeted offers.

Explore data-driven segmentation with decision trees and pattern aggregation to reveal segments like divorced customers with kids and income-based bins for targeted boat rental marketing.

Prioritize real problems solved with a mentor, and learn to code in R or Python to apply data science and analytics skills to business challenges.