{"id":3471,"date":"2023-08-31T11:12:59","date_gmt":"2023-08-31T11:12:59","guid":{"rendered":"https:\/\/neuraldesigner.com\/learning\/car-co2-emissions\/"},"modified":"2026-08-05T14:58:30","modified_gmt":"2026-08-05T12:58:30","slug":"car-co2-emissions","status":"publish","type":"learning","link":"https:\/\/www.neuraldesigner.com\/learning\/examples\/car-co2-emissions\/","title":{"rendered":"Estimate car emissions using machine learning"},"content":{"rendered":"<style>\n.ndb{width:100vw;margin-left:calc(50% - 50vw);padding:22px 24px 14px;background:#eee;color:#1b2635;font-family:\"Outfit\",\"Roboto\",Arial,sans-serif}\n.ndb *{box-sizing:border-box}.ndb-wrap{width:min(100%,1200px);margin:0 auto}.ndb a{text-decoration:none}\n.ndb-executive{margin:0 0 34px;padding:32px;border-radius:20px;background:linear-gradient(135deg,#12354b,#245e80);color:#fff}\n.ndb-executive h2{margin:0 0 12px;color:#fff;font-size:30px}.ndb-executive p{font-size:18px;line-height:1.55}\n.ndb-kpis{display:grid;grid-template-columns:repeat(4,minmax(0,1fr));gap:14px;margin-top:22px}.ndb-kpi{padding:18px;border:1px solid rgba(255,255,255,.18);border-radius:14px;background:rgba(255,255,255,.1)}.ndb-kpi strong{display:block;font-size:25px}.ndb-kpi span{font-size:13px}\n.ndb-actions,.ndb-audience,.ndb-downloads{display:flex;flex-wrap:wrap;justify-content:center;gap:12px;margin:22px 0}.ndb-actions a,.ndb-downloads a{padding:12px 20px;border-radius:24px;background:#245e80;color:#fff;font-weight:700}.ndb-executive .ndb-actions a{background:#fff;color:#12354b}\n.ndb-toc{display:flex;flex-wrap:wrap;justify-content:center;gap:10px;margin:0 0 42px;padding:0;list-style:none}.ndb-toc a,.ndb-audience span{display:block;padding:9px 14px;border-radius:20px;background:#e9f2f8;color:#12354b;font-weight:600}\n.ndb-card{margin:0 0 54px;scroll-margin-top:90px}.ndb-card h2{margin:0 0 20px;padding-bottom:12px;border-bottom:1px solid #dbe5ec;color:#001233;font-size:24px}.ndb-card p,.ndb-card li{font-size:16.5px;line-height:1.6}.ndb-card img{display:block;width:auto;max-width:min(640px,100%);height:auto;margin:24px auto;border-radius:12px}.ndb-card img.ndb-architecture{width:min(1000px,100%);max-width:100%}\n.ndb-value-grid{display:grid;grid-template-columns:repeat(3,minmax(0,1fr));gap:16px;margin:22px 0}.ndb-value{padding:20px;border:1px solid #dce8ef;border-radius:14px;background:#f8fbfd}\n.ndb-note{margin:20px 0;padding:18px 20px;border-left:4px solid #56a1c8;border-radius:0 12px 12px 0;background:#f6fafc}\n.ndb-card table{width:auto;max-width:100%;margin:24px auto;border-collapse:collapse;background:#fff}.ndb-card th,.ndb-card td{padding:11px 16px;border-bottom:1px solid #e2e9ee}.ndb-card thead th{background:#12354b!important;color:#fff!important}.ndb-card tbody th{background:#eaf2f6!important;color:#12354b!important;text-align:left}.ndb-card tbody tr:nth-child(even) th{background:#f4f8fa!important}.ndb-card tbody td{color:#33424f!important}\n.ndb-figure-grid{display:grid;grid-template-columns:repeat(2,minmax(0,1fr));gap:20px}.ndb-figure-grid figure{margin:0;padding:16px;border:1px solid #dce8ef;border-radius:15px;background:#f8fbfd}.ndb-figure-grid img{width:100%;max-width:100%;margin:0 auto 12px}\n@media(max-width:820px){.ndb-kpis{grid-template-columns:repeat(2,minmax(0,1fr))}.ndb-value-grid{grid-template-columns:1fr}}\n@media(max-width:620px){.ndb{padding:12px 14px}.ndb-kpis,.ndb-figure-grid{grid-template-columns:1fr}.ndb-executive{padding:24px 20px}}\n<\/style>\n<style>\n.ndb{box-sizing:border-box;overflow-x:hidden}.ndb a{color:#2d799f}.ndb-executive a{color:inherit}\n.ndb-wrap{line-height:1.45}.ndb-lead{margin:0 0 28px;color:#3a4a5a;font-size:18px;line-height:1.6}\n.ndb-card h2{position:relative}.ndb-card h2:after{content:\"\";position:absolute;bottom:-1px;left:0;width:62px;height:3px;border-radius:2px;background:#56a1c8}\n.ndb-card h3{margin:28px 0 12px;color:#12354b;font-size:20px}.ndb-card code{padding:2px 5px;border-radius:4px;background:#e8eef2;color:#12354b}\n.ndb-card img{box-shadow:0 12px 28px rgba(0,18,51,.12)}.ndb-card img.ndb-architecture{width:min(1000px,100%)}.ndb-chart-wide{width:min(780px,100%)!important;max-width:100%!important}.ndb-chart-square{width:min(540px,100%)!important;max-width:100%!important}\n.ndb-figure-grid{margin:24px 0}.ndb-figure-grid figure{display:flex;flex-direction:column;justify-content:space-between}.ndb-figure-grid img{box-shadow:none}.ndb-figure-grid figcaption{color:#405361;font-size:14px;line-height:1.45}.ndb-figure-full{grid-column:1\/-1}.ndb-figure-full img{width:min(540px,100%)!important}\n.ndb-table-scroll{max-width:100%;overflow-x:auto}.ndb-card table{border-collapse:separate;border-spacing:0;overflow:hidden;border-radius:12px;box-shadow:0 10px 24px rgba(0,18,51,.08)}.ndb-card tbody tr:last-child td{border-bottom:0}.ndb-card table.ndb-metrics{font-variant-numeric:tabular-nums}.ndb-card table.ndb-metrics td:not(:first-child){text-align:right}\n.ndb-card .ndb-kpi{border-color:#dce8ef;background:#f8fbfd}.ndb-card .ndb-kpi strong{color:#12354b}.ndb-card .ndb-kpi span{color:#5e707d}\n.ndb-flow{display:grid;grid-template-columns:repeat(4,minmax(0,1fr));gap:12px;margin:24px 0}.ndb-flow div{position:relative;display:flex;min-height:88px;align-items:center;justify-content:center;padding:18px 16px;border-radius:13px;background:#12354b;color:#fff;text-align:center;font-weight:700}.ndb-flow div:not(:last-child):after{content:\"\u2192\";position:absolute;right:-17px;top:50%;z-index:2;transform:translateY(-50%);color:#56a1c8;font-size:22px}\n.ndb-case{margin:26px 0;padding:24px;border:1px solid #cfe0ea;border-radius:16px;background:#f8fbfd}.ndb-case h3{margin-top:0}.ndb-case-grid{display:grid;grid-template-columns:repeat(2,minmax(0,1fr));gap:20px}.ndb-case-grid table{width:100%!important;margin:0 auto!important}\n.ndb-note--warning{border-left-color:#d79a29;background:#fff9ed}.ndb-note--critical{border-left-color:#c64c4c;background:#fff5f5}\n@media(max-width:820px){.ndb-flow{grid-template-columns:1fr 1fr}.ndb-flow div:after{display:none}.ndb-case-grid{grid-template-columns:1fr}}\n@media(max-width:620px){.ndb-flow{grid-template-columns:1fr}.ndb-card table{font-size:13px}.ndb-card th,.ndb-card td{padding:9px 8px}}\n<\/style>\n\n<div class=\"ndb\">\n<div class=\"ndb-wrap\">\n<section class=\"ndb-executive\">\n<h2>Estimate vehicle CO\u2082 ratings from fuel-consumption data<\/h2>\n<p>This example learns the relationship between vehicle specifications, standardized fuel-consumption ratings and reported carbon-dioxide emissions. It can support catalogue screening, engineering comparison and data-quality checks while keeping certification decisions with approved test procedures.<\/p>\n<div class=\"ndb-kpis\">\n<div class=\"ndb-kpi\"><strong>7,385<\/strong><span>vehicle records<\/span><\/div>\n<div class=\"ndb-kpi\"><strong>11<\/strong><span>logical inputs<\/span><\/div>\n<div class=\"ndb-kpi\"><strong>0.9945<\/strong><span>testing determination<\/span><\/div>\n<div class=\"ndb-kpi\"><strong>2.47 g\/km<\/strong><span>testing MAE<\/span><\/div>\n<\/div>\n<div class=\"ndb-actions\"><a href=\"#7-model-deployment\">Review the deployment case<\/a><a href=\"https:\/\/www.neuraldesigner.com\/wp-content\/uploads\/2026\/08\/co2_emissions.csv\">Download the data<\/a><\/div>\n<\/section>\n<ul class=\"ndb-toc\">\n<li><a href=\"#1-industrial-challenge\">Industrial challenge<\/a><\/li><li><a href=\"#2-data-set\">Data set<\/a><\/li><li><a href=\"#3-model\">Model<\/a><\/li><li><a href=\"#4-training\">Training<\/a><\/li><li><a href=\"#5-selection\">Model selection<\/a><\/li><li><a href=\"#6-testing\">Testing<\/a><\/li><li><a href=\"#7-model-deployment\">Deployment<\/a><\/li><li><a href=\"#8-limitations\">Limitations<\/a><\/li>\n<\/ul>\n<section id=\"1-industrial-challenge\" class=\"ndb-card\"><h2>1. Industrial challenge<\/h2><p>Vehicle programmes, fleet teams and sustainability functions often need an early estimate or an independent plausibility check before a complete certification record is available. This approximation model maps categorical vehicle descriptors and standardized fuel-consumption ratings to estimated <code>co2_emissions<\/code> in grams per kilometre.<\/p><div class=\"ndb-value-grid\"><div class=\"ndb-value\"><strong>Screen vehicle specifications<\/strong><span>Estimate whether a catalogue CO\u2082 value is consistent with its powertrain and fuel-consumption record.<\/span><\/div><div class=\"ndb-value\"><strong>Compare fleet candidates<\/strong><span>Use one consistent analytical workflow to support procurement and portfolio decarbonization reviews.<\/span><\/div><div class=\"ndb-value\"><strong>Prioritize data-quality checks<\/strong><span>Flag large measured-versus-expected differences for source-data or homologation review.<\/span><\/div><\/div><div class=\"ndb-audience\"><span>Vehicle engineering<\/span><span>Homologation<\/span><span>Fleet management<\/span><span>Sustainability<\/span><span>Automotive data quality<\/span><\/div><div class=\"ndb-note\"><strong>Scope of this example.<\/strong> The target is a model-specific fuel-consumption rating expressed as g CO\u2082\/km. It is not an on-road emissions measurement, regulatory certificate or substitute for an approved vehicle test.<\/div><\/section>\n<section id=\"2-data-set\" class=\"ndb-card\"><h2>2. Data set<\/h2>\n<p>The updated <a href=\"https:\/\/www.neuraldesigner.com\/wp-content\/uploads\/2026\/08\/co2_emissions.csv\"><code>co2_emissions.csv<\/code><\/a> contains <strong>7,385 records<\/strong>, eleven inputs and one target. The prepared file uses semicolon delimiters and contains no missing values.<\/p>\n<div class=\"ndb-downloads\"><a href=\"https:\/\/www.neuraldesigner.com\/wp-content\/uploads\/2026\/08\/co2_emissions.csv\">Download co2_emissions.csv<\/a><\/div>\n<div class=\"ndb-table-scroll\"><table><thead><tr><th>Variable<\/th><th>Meaning<\/th><th>Type, unit or range<\/th><\/tr><\/thead><tbody>\n<tr><td><code>brand<\/code><\/td><td>Vehicle manufacturer<\/td><td>Categorical; 42 brands<\/td><\/tr>\n<tr><td><code>model<\/code><\/td><td>Commercial model designation<\/td><td>Categorical; 2,053 values<\/td><\/tr>\n<tr><td><code>vehicle_class<\/code><\/td><td>Vehicle body\/class category<\/td><td>Categorical; 16 classes<\/td><\/tr>\n<tr><td><code>engine_size<\/code><\/td><td>Engine displacement<\/td><td>0.9 to 8.4 L<\/td><\/tr>\n<tr><td><code>cylinders<\/code><\/td><td>Engine cylinder count<\/td><td>3 to 16<\/td><\/tr>\n<tr><td><code>transmission<\/code><\/td><td>A automatic, AM automated manual, AS automatic select shift, AV continuously variable, M manual<\/td><td>5 categories<\/td><\/tr>\n<tr><td><code>fuel_type<\/code><\/td><td>D diesel, E ethanol\/E85, N natural gas, X regular gasoline, Z premium gasoline<\/td><td>5 categories<\/td><\/tr>\n<tr><td><code>fuel_consumption_city<\/code><\/td><td>Standardized city fuel consumption<\/td><td>4.2 to 30.6 L\/100 km<\/td><\/tr>\n<tr><td><code>fuel_consumption_hwy<\/code><\/td><td>Standardized highway fuel consumption<\/td><td>4.0 to 20.6 L\/100 km<\/td><\/tr>\n<tr><td><code>fuel_consumption_comb(l\/100km)<\/code><\/td><td>Combined fuel consumption<\/td><td>4.1 to 26.1 L\/100 km<\/td><\/tr>\n<tr><td><code>fuel_consumption_comb(mpg)<\/code><\/td><td>Combined fuel economy<\/td><td>11 to 69 mpg<\/td><\/tr>\n<tr><td><code>co2_emissions<\/code><\/td><td>Reported CO\u2082 emissions rating<\/td><td>Target; 96 to 522 g\/km<\/td><\/tr>\n<\/tbody><\/table><\/div>\n<div class=\"ndb-figure-grid\"><figure><img decoding=\"async\" src=\"https:\/\/www.neuraldesigner.com\/wp-content\/uploads\/2026\/08\/car-co2-distribution-2026.png\" alt=\"Distribution of vehicle carbon-dioxide emissions in grams per kilometre\"><figcaption><strong>Target distribution.<\/strong> Most records lie between approximately 180 and 330 g\/km, with a sparse high-emission tail.<\/figcaption><\/figure><figure><img decoding=\"async\" src=\"https:\/\/www.neuraldesigner.com\/wp-content\/uploads\/2026\/08\/car-co2-correlations-2026.png\" alt=\"Pearson correlations between vehicle inputs and carbon-dioxide emissions\"><figcaption><strong>Dominant numeric drivers.<\/strong> Combined, city and highway consumption have the strongest positive relationships; combined mpg has the expected inverse relationship. Correlations for encoded categorical variables are not an interpretable importance measure.<\/figcaption><\/figure><\/div>\n<img decoding=\"async\" class=\"ndb-chart-wide\" src=\"https:\/\/www.neuraldesigner.com\/wp-content\/uploads\/2026\/08\/car-co2-mpg-scatter-2026.png\" alt=\"Vehicle carbon-dioxide emissions against combined fuel economy in miles per gallon\">\n<p>The curved inverse pattern against mpg is expected because mpg and L\/100 km are reciprocal representations of fuel economy. Including both variables gives the network redundant information and should be reviewed in a production feature set.<\/p>\n<div class=\"ndb-note\"><strong>Validation design.<\/strong> The saved project uses a random 60\/20\/20 split: 4,431 training, 1,477 selection and 1,477 testing rows. The file contains 1,112 exact duplicate rows; 293 testing rows also appear exactly in training. Grouping duplicates and restoring model year before splitting would provide a more defensible estimate for unseen vehicles.<\/div><\/section>\n<section id=\"3-model\" class=\"ndb-card\"><h2>3. Model<\/h2><p>The network receives eleven logical variables. One-hot encoding expands the five categorical inputs\u2014especially the 2,053-value <code>model<\/code> field\u2014to <strong>2,127 input features<\/strong>. Numeric variables use mean-and-standard-deviation scaling, while categorical features use minimum\u2013maximum scaling.<\/p><p>A hidden layer with three tanh neurons feeds one identity output neuron. The result is unscaled to g CO\u2082\/km. The saved project applies no output bounding.<\/p><img decoding=\"async\" class=\"ndb-architecture\" src=\"https:\/\/www.neuraldesigner.com\/wp-content\/uploads\/2026\/08\/car-co2-network-2026.png\" alt=\"Vehicle CO2 neural network with eleven logical inputs, three hidden neurons and one output\"><\/section>\n<section id=\"4-training\" class=\"ndb-card\"><h2>4. Training strategy<\/h2>\n<p>The model minimizes normalized squared error with L2 regularization of 0.01 and the quasi-Newton method. The recorded run contains 152 epochs: training error falls from 0.974 to 0.00559, while selection error falls from 0.180 to 0.00135.<\/p>\n<img decoding=\"async\" class=\"ndb-chart-wide\" src=\"https:\/\/www.neuraldesigner.com\/wp-content\/uploads\/2026\/08\/car-co2-training-history-2026.png\" alt=\"Quasi-Newton training and selection error history for vehicle CO2 prediction\">\n<p>The smooth convergence supports optimization of the configured network. The unusually low selection error should still be interpreted alongside duplicate overlap and the strong deterministic relationship between fuel consumption, fuel type and CO\u2082 rating.<\/p>\n<\/section>\n<section id=\"5-selection\" class=\"ndb-card\"><h2>5. Model selection<\/h2><p>No separate neuron-selection or input-selection task is stored in the regenerated output. The example therefore keeps the configured three-neuron hidden layer and does not claim it as an independently selected optimum.<\/p><div class=\"ndb-note ndb-note--warning\"><strong>Professional next step.<\/strong> Repeat selection after grouping duplicate records and compare an engineering-only feature set that excludes high-cardinality <code>model<\/code> identifiers and one of the two equivalent combined-consumption measures.<\/div><\/section>\n<section id=\"6-testing\" class=\"ndb-card\"><h2>6. Testing analysis<\/h2><p>The final network is evaluated on 1,477 testing records. Neural Designer reports a determination coefficient of 0.9945. The conventional SSE-based R\u00b2 is 0.9944.<\/p>\n<div class=\"ndb-kpis\"><div class=\"ndb-kpi\"><strong>0.9945<\/strong><span>determination<\/span><\/div><div class=\"ndb-kpi\"><strong>2.47 g\/km<\/strong><span>mean absolute error<\/span><\/div><div class=\"ndb-kpi\"><strong>4.36 g\/km<\/strong><span>root mean squared error<\/span><\/div><div class=\"ndb-kpi\"><strong>6.32 g\/km<\/strong><span>95th percentile absolute error<\/span><\/div><\/div>\n<h3>Transparent baseline comparison<\/h3>\n<p>A linear benchmark was fitted on the same training rows using only combined L\/100 km and fuel-type indicators, then evaluated on the same testing rows.<\/p>\n<table class=\"ndb-metrics\"><thead><tr><th>Model<\/th><th>Testing R\u00b2<\/th><th>MAE (g\/km)<\/th><th>RMSE (g\/km)<\/th><\/tr><\/thead><tbody><tr><td>Neural network<\/td><td>0.9944<\/td><td>2.47<\/td><td>4.36<\/td><\/tr><tr><td>Linear consumption + fuel baseline<\/td><td>0.9921<\/td><td>3.06<\/td><td>5.20<\/td><\/tr><\/tbody><\/table>\n<p>The network reduces MAE by 19.2% relative to this strong transparent baseline. That improvement is real on the configured split, but it is modest enough that interpretability and unseen-category handling should influence the production choice.<\/p>\n<div class=\"ndb-figure-grid\">\n<figure class=\"ndb-figure-full\"><img decoding=\"async\" class=\"ndb-chart-square\" src=\"https:\/\/www.neuraldesigner.com\/wp-content\/uploads\/2026\/08\/car-co2-gof-2026.png\" alt=\"Observed versus predicted vehicle carbon-dioxide emissions on the testing set\"><figcaption><strong>Goodness of fit.<\/strong> Most predictions follow the identity line closely. The largest absolute error is 87.0 g\/km, and the highest observed value is visibly underpredicted.<\/figcaption><\/figure>\n<\/div><\/section>\n<section id=\"7-model-deployment\" class=\"ndb-card\"><h2>7. Model deployment<\/h2>\n<p>A defensible application is a catalogue or homologation-data consistency check, not a replacement for the official test. Vehicle data is validated, passed to the model, compared with the recorded rating and routed for review when the residual exceeds a documented tolerance.<\/p>\n<div class=\"ndb-flow\"><div>Vehicle specification and rating data<\/div><div>Schema, category and unit validation<\/div><div>CO\u2082 estimation model<\/div><div>Residual check and engineering review<\/div><\/div>\n<div class=\"ndb-case\"><h3>Representative testing case: Toyota Camry<\/h3>\n<p>This input combination is a real testing record and does not occur in the training subset. Its absolute error is close to the overall testing MAE, making it more representative than a specially selected near-perfect prediction.<\/p>\n<div class=\"ndb-case-grid\"><table><thead><tr><th>Vehicle input<\/th><th>Value<\/th><\/tr><\/thead><tbody>\n<tr><td>Brand \/ model<\/td><td>TOYOTA CAMRY<\/td><\/tr><tr><td>Vehicle class<\/td><td>Mid-size<\/td><\/tr><tr><td>Engine<\/td><td>2.5 L, 4 cylinders<\/td><\/tr><tr><td>Transmission<\/td><td>AS \u2014 automatic select shift<\/td><\/tr><tr><td>Fuel<\/td><td>X \u2014 regular gasoline<\/td><\/tr><tr><td>City \/ highway<\/td><td>9.5 \/ 6.6 L\/100 km<\/td><\/tr><tr><td>Combined consumption<\/td><td>8.2 L\/100 km (34 mpg)<\/td><\/tr>\n<\/tbody><\/table><table><thead><tr><th>Result<\/th><th>g CO\u2082\/km<\/th><\/tr><\/thead><tbody><tr><td>Reported rating<\/td><td>189<\/td><\/tr><tr><td>Model estimate<\/td><td>191.1<\/td><\/tr><tr><td>Residual<\/td><td>+2.1 (+1.1%)<\/td><\/tr><tr><td>Empirical p95 absolute-error reference<\/td><td>\u00b16.3<\/td><\/tr><\/tbody><\/table><\/div>\n<p><strong>Operational interpretation:<\/strong> the estimate is 2.1 g\/km from the recorded value, inside the model&#8217;s empirical 95th-percentile absolute-error reference. A workflow could accept this row as internally consistent while sending larger residuals for source-data and engineering review. The p95 value is a screening tolerance reference, not a formal prediction interval.<\/p><\/div>\n<h3>Local sensitivity at a real vehicle reference<\/h3>\n<p>The regenerated directional-output task now starts from a valid vehicle record. The HTML table reproduces the reference values used by Neural Designer.<\/p>\n<table><thead><tr><th>Reference input<\/th><th>Value<\/th><\/tr><\/thead><tbody>\n<tr><td>Vehicle<\/td><td>ACURA RLX HYBRID, mid-size<\/td><\/tr><tr><td>Engine<\/td><td>3.5 L, 6 cylinders<\/td><\/tr><tr><td>Transmission \/ fuel<\/td><td>AM \/ Z \u2014 automated manual \/ premium gasoline<\/td><\/tr><tr><td>City consumption<\/td><td>8.0 L\/100 km<\/td><\/tr><tr><td>Highway consumption<\/td><td>7.5 L\/100 km<\/td><\/tr><tr><td>Combined consumption<\/td><td>7.7 L\/100 km (37 mpg)<\/td><\/tr><tr><td>Model estimate at reference<\/td><td>181.1 g CO\u2082\/km<\/td><\/tr>\n<\/tbody><\/table>\n<img decoding=\"async\" class=\"ndb-chart-wide\" src=\"https:\/\/www.neuraldesigner.com\/wp-content\/uploads\/2026\/08\/car-co2-city-consumption-directional-2026.png\" alt=\"Local sensitivity of predicted vehicle carbon-dioxide emissions to city fuel consumption from an ACURA RLX HYBRID reference point\">\n<p>The curve shows the model response when only <code>fuel_consumption_city<\/code> changes and every other input remains fixed. Interpret it only inside the observed city-consumption range of <strong>4.2 to 30.6 L\/100 km<\/strong>; the chart extends beyond that domain for visualization. Because city, highway, combined L\/100 km and mpg are mathematically related, this is a local sensitivity analysis rather than a feasible vehicle-configuration scenario. Operational comparisons should update the linked consumption fields consistently or compare complete observed configurations.<\/p>\n<div class=\"ndb-note ndb-note--critical\"><strong>Deployment boundary.<\/strong> Validate category mappings, reject unseen brands\/models or route them to a fallback model, version every prediction, monitor residual and feature drift, and keep regulatory or procurement decisions under qualified human review. This model is not a certified emissions or homologation system.<\/div>\n<div class=\"ndb-downloads\"><a href=\"https:\/\/www.neuraldesigner.com\/wp-content\/uploads\/2026\/08\/co2_emissions.csv\">Download dataset (CSV)<\/a><\/div><\/section>\n<section id=\"8-limitations\" class=\"ndb-card\"><h2>8. Scope and limitations<\/h2>\n<ul>\n<li>The prepared extract does not contain model year, so temporal validation and year-to-year regulatory or powertrain shifts cannot be tested.<\/li>\n<li>There are 1,112 exact duplicate rows. Random splitting places 293 exact testing rows in training and 462 testing input combinations in training.<\/li>\n<li>The <code>model<\/code> field has 2,053 categories. New vehicle names require an explicit unknown-category policy or a model based on transferable engineering variables.<\/li>\n<li>City, highway and combined consumption are strongly related, while combined mpg is a reciprocal representation of combined L\/100 km. This multicollinearity increases redundancy.<\/li>\n<li>Fuel consumption and fuel type already determine most of the reported CO\u2082 rating. The neural network should be compared with a transparent emissions-factor or linear calculation before deployment.<\/li>\n<li>The target represents standardized\/model-specific ratings, not real-world driving emissions. Driver behaviour, load, weather, traffic, maintenance and route conditions are outside the model.<\/li>\n<li>The highest-emission observations are sparse, and the goodness-of-fit chart shows underprediction at the extreme tail.<\/li>\n<li>Production validation should hold out complete model families and future model years, document unit and category mappings, and monitor data and residual drift.<\/li>\n<\/ul>\n<\/section>\n<section id=\"references\" class=\"ndb-card\"><h2>References<\/h2>\n<ul><li>Natural Resources Canada. <a href=\"https:\/\/open.canada.ca\/data\/en\/dataset\/98f1a129-f628-4ce4-b24d-6f16bf24dd64\" target=\"_blank\" rel=\"noopener\">Fuel consumption ratings<\/a>: official context and variable units for model-specific fuel-consumption ratings and estimated CO\u2082 emissions.<\/li><li>Kaggle. <a href=\"https:\/\/www.kaggle.com\/datasets\/debajyotipodder\/co2-emission-by-vehicles\" target=\"_blank\" rel=\"noopener\">CO\u2082 Emission by Vehicles<\/a>: prepared dataset used by this example.<\/li><li><a href=\"https:\/\/www.neuraldesigner.com\/learning\/tutorials\/testing-analysis\/\">Neural Designer testing analysis<\/a>: goodness-of-fit and error interpretation.<\/li><li><a href=\"https:\/\/www.neuraldesigner.com\/learning\/tutorials\/model-deployment\/\">Neural Designer model deployment<\/a>: directional outputs and deployment concepts.<\/li><\/ul>\n<\/section>\n<\/div>\n<\/div>\n","protected":false},"author":11,"featured_media":2475,"template":"","categories":[29],"tags":[41,46],"class_list":["post-3471","learning","type-learning","status-publish","has-post-thumbnail","hentry","category-examples","tag-automotive","tag-environment"],"acf":[],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v26.4 - https:\/\/yoast.com\/wordpress\/plugins\/seo\/ -->\n<title>Estimate car emissions using machine learning<\/title>\n<meta name=\"description\" content=\"Build a machine learning model to predict CO2 emissions from car features such as engine size, transmission type, etc.\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/www.neuraldesigner.com\/learning\/examples\/car-co2-emissions\/\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Car CO2 emissions machine learning example\" \/>\n<meta property=\"og:description\" content=\"Nowadays, it is very important to reduce CO2 emissions in the atmosphere in order to fight against climate change. Use Neural Designer to estimate CO2 emissions from cars.\" \/>\n<meta property=\"og:url\" content=\"https:\/\/www.neuraldesigner.com\/learning\/examples\/car-co2-emissions\/\" \/>\n<meta property=\"og:site_name\" content=\"Neural Designer\" \/>\n<meta property=\"article:modified_time\" content=\"2026-08-05T12:58:30+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/www.neuraldesigner.com\/wp-content\/uploads\/2023\/06\/car-road-co2-emissions.webp\" \/>\n\t<meta property=\"og:image:width\" content=\"1200\" \/>\n\t<meta property=\"og:image:height\" content=\"628\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/webp\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:title\" content=\"Car CO2 emissions machine learning example\" \/>\n<meta name=\"twitter:description\" content=\"Nowadays, it is very important to reduce CO2 emissions in the atmosphere in order to fight against climate change. Use Neural Designer to estimate CO2 emissions from cars.\" \/>\n<meta name=\"twitter:image\" content=\"https:\/\/www.neuraldesigner.com\/wp-content\/uploads\/2023\/06\/car-road-co2-emissions.webp\" \/>\n<meta name=\"twitter:site\" content=\"@NeuralDesigner\" \/>\n<meta name=\"twitter:label1\" content=\"Est. reading time\" \/>\n\t<meta name=\"twitter:data1\" content=\"8 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\/\/www.neuraldesigner.com\/learning\/examples\/car-co2-emissions\/\",\"url\":\"https:\/\/www.neuraldesigner.com\/learning\/examples\/car-co2-emissions\/\",\"name\":\"Estimate car emissions using machine learning\",\"isPartOf\":{\"@id\":\"https:\/\/www.neuraldesigner.com\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\/\/www.neuraldesigner.com\/learning\/examples\/car-co2-emissions\/#primaryimage\"},\"image\":{\"@id\":\"https:\/\/www.neuraldesigner.com\/learning\/examples\/car-co2-emissions\/#primaryimage\"},\"thumbnailUrl\":\"https:\/\/www.neuraldesigner.com\/wp-content\/uploads\/2023\/06\/car-road-co2-emissions.webp\",\"datePublished\":\"2023-08-31T11:12:59+00:00\",\"dateModified\":\"2026-08-05T12:58:30+00:00\",\"description\":\"Build a machine learning model to predict CO2 emissions from car features such as engine size, transmission type, etc.\",\"breadcrumb\":{\"@id\":\"https:\/\/www.neuraldesigner.com\/learning\/examples\/car-co2-emissions\/#breadcrumb\"},\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/www.neuraldesigner.com\/learning\/examples\/car-co2-emissions\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\/\/www.neuraldesigner.com\/learning\/examples\/car-co2-emissions\/#primaryimage\",\"url\":\"https:\/\/www.neuraldesigner.com\/wp-content\/uploads\/2023\/06\/car-road-co2-emissions.webp\",\"contentUrl\":\"https:\/\/www.neuraldesigner.com\/wp-content\/uploads\/2023\/06\/car-road-co2-emissions.webp\",\"width\":1200,\"height\":628,\"caption\":\"Car CO2 emissions on the road\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/www.neuraldesigner.com\/learning\/examples\/car-co2-emissions\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\/\/www.neuraldesigner.com\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Learning\",\"item\":\"https:\/\/www.neuraldesigner.com\/learning\/\"},{\"@type\":\"ListItem\",\"position\":3,\"name\":\"Estimate car emissions using machine learning\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/www.neuraldesigner.com\/#website\",\"url\":\"https:\/\/www.neuraldesigner.com\/\",\"name\":\"Neural Designer\",\"description\":\"Explanable AI Platform\",\"publisher\":{\"@id\":\"https:\/\/www.neuraldesigner.com\/#organization\"},\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/www.neuraldesigner.com\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"en-US\"},{\"@type\":\"Organization\",\"@id\":\"https:\/\/www.neuraldesigner.com\/#organization\",\"name\":\"Neural Designer\",\"url\":\"https:\/\/www.neuraldesigner.com\/\",\"logo\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\/\/www.neuraldesigner.com\/#\/schema\/logo\/image\/\",\"url\":\"https:\/\/www.neuraldesigner.com\/wp-content\/uploads\/2023\/05\/logo-neural-1.png\",\"contentUrl\":\"https:\/\/www.neuraldesigner.com\/wp-content\/uploads\/2023\/05\/logo-neural-1.png\",\"width\":1024,\"height\":223,\"caption\":\"Neural Designer\"},\"image\":{\"@id\":\"https:\/\/www.neuraldesigner.com\/#\/schema\/logo\/image\/\"},\"sameAs\":[\"https:\/\/x.com\/NeuralDesigner\",\"https:\/\/es.linkedin.com\/showcase\/neuraldesigner\/\"]}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"Estimate car emissions using machine learning","description":"Build a machine learning model to predict CO2 emissions from car features such as engine size, transmission type, etc.","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/www.neuraldesigner.com\/learning\/examples\/car-co2-emissions\/","og_locale":"en_US","og_type":"article","og_title":"Car CO2 emissions machine learning example","og_description":"Nowadays, it is very important to reduce CO2 emissions in the atmosphere in order to fight against climate change. Use Neural Designer to estimate CO2 emissions from cars.","og_url":"https:\/\/www.neuraldesigner.com\/learning\/examples\/car-co2-emissions\/","og_site_name":"Neural Designer","article_modified_time":"2026-08-05T12:58:30+00:00","og_image":[{"width":1200,"height":628,"url":"https:\/\/www.neuraldesigner.com\/wp-content\/uploads\/2023\/06\/car-road-co2-emissions.webp","type":"image\/webp"}],"twitter_card":"summary_large_image","twitter_title":"Car CO2 emissions machine learning example","twitter_description":"Nowadays, it is very important to reduce CO2 emissions in the atmosphere in order to fight against climate change. Use Neural Designer to estimate CO2 emissions from cars.","twitter_image":"https:\/\/www.neuraldesigner.com\/wp-content\/uploads\/2023\/06\/car-road-co2-emissions.webp","twitter_site":"@NeuralDesigner","twitter_misc":{"Est. reading time":"8 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/www.neuraldesigner.com\/learning\/examples\/car-co2-emissions\/","url":"https:\/\/www.neuraldesigner.com\/learning\/examples\/car-co2-emissions\/","name":"Estimate car emissions using machine learning","isPartOf":{"@id":"https:\/\/www.neuraldesigner.com\/#website"},"primaryImageOfPage":{"@id":"https:\/\/www.neuraldesigner.com\/learning\/examples\/car-co2-emissions\/#primaryimage"},"image":{"@id":"https:\/\/www.neuraldesigner.com\/learning\/examples\/car-co2-emissions\/#primaryimage"},"thumbnailUrl":"https:\/\/www.neuraldesigner.com\/wp-content\/uploads\/2023\/06\/car-road-co2-emissions.webp","datePublished":"2023-08-31T11:12:59+00:00","dateModified":"2026-08-05T12:58:30+00:00","description":"Build a machine learning model to predict CO2 emissions from car features such as engine size, transmission type, etc.","breadcrumb":{"@id":"https:\/\/www.neuraldesigner.com\/learning\/examples\/car-co2-emissions\/#breadcrumb"},"inLanguage":"en-US","potentialAction":[{"@type":"ReadAction","target":["https:\/\/www.neuraldesigner.com\/learning\/examples\/car-co2-emissions\/"]}]},{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/www.neuraldesigner.com\/learning\/examples\/car-co2-emissions\/#primaryimage","url":"https:\/\/www.neuraldesigner.com\/wp-content\/uploads\/2023\/06\/car-road-co2-emissions.webp","contentUrl":"https:\/\/www.neuraldesigner.com\/wp-content\/uploads\/2023\/06\/car-road-co2-emissions.webp","width":1200,"height":628,"caption":"Car CO2 emissions on the road"},{"@type":"BreadcrumbList","@id":"https:\/\/www.neuraldesigner.com\/learning\/examples\/car-co2-emissions\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/www.neuraldesigner.com\/"},{"@type":"ListItem","position":2,"name":"Learning","item":"https:\/\/www.neuraldesigner.com\/learning\/"},{"@type":"ListItem","position":3,"name":"Estimate car emissions using machine learning"}]},{"@type":"WebSite","@id":"https:\/\/www.neuraldesigner.com\/#website","url":"https:\/\/www.neuraldesigner.com\/","name":"Neural Designer","description":"Explanable AI Platform","publisher":{"@id":"https:\/\/www.neuraldesigner.com\/#organization"},"potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/www.neuraldesigner.com\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"en-US"},{"@type":"Organization","@id":"https:\/\/www.neuraldesigner.com\/#organization","name":"Neural Designer","url":"https:\/\/www.neuraldesigner.com\/","logo":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/www.neuraldesigner.com\/#\/schema\/logo\/image\/","url":"https:\/\/www.neuraldesigner.com\/wp-content\/uploads\/2023\/05\/logo-neural-1.png","contentUrl":"https:\/\/www.neuraldesigner.com\/wp-content\/uploads\/2023\/05\/logo-neural-1.png","width":1024,"height":223,"caption":"Neural Designer"},"image":{"@id":"https:\/\/www.neuraldesigner.com\/#\/schema\/logo\/image\/"},"sameAs":["https:\/\/x.com\/NeuralDesigner","https:\/\/es.linkedin.com\/showcase\/neuraldesigner\/"]}]}},"_links":{"self":[{"href":"https:\/\/www.neuraldesigner.com\/api\/wp\/v2\/learning\/3471","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.neuraldesigner.com\/api\/wp\/v2\/learning"}],"about":[{"href":"https:\/\/www.neuraldesigner.com\/api\/wp\/v2\/types\/learning"}],"author":[{"embeddable":true,"href":"https:\/\/www.neuraldesigner.com\/api\/wp\/v2\/users\/11"}],"version-history":[{"count":8,"href":"https:\/\/www.neuraldesigner.com\/api\/wp\/v2\/learning\/3471\/revisions"}],"predecessor-version":[{"id":23070,"href":"https:\/\/www.neuraldesigner.com\/api\/wp\/v2\/learning\/3471\/revisions\/23070"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.neuraldesigner.com\/api\/wp\/v2\/media\/2475"}],"wp:attachment":[{"href":"https:\/\/www.neuraldesigner.com\/api\/wp\/v2\/media?parent=3471"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.neuraldesigner.com\/api\/wp\/v2\/categories?post=3471"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.neuraldesigner.com\/api\/wp\/v2\/tags?post=3471"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}