{ "fetched_at": "2026-10-19T13:35:58Z", "source_url": "https://www.kaggle.com/competitions/1dl034-project-vt24/data", "pages": [ { "name": "data-description", "content": "# The Forest CoverType dataset\n\n1. Title of Database:\nForest Covertype data\n\n2. Sources:\n (a) Original owners of database:\n\tRemote Sensing and GIS Program\n\tDepartment of Forest Sciences\n\tCollege of Natural Resources\n\tColorado State University\n\tFort Collins, CO 80523\n\t(contact Jock A. Blackard, jblackard 'at' fs.fed.us or Dr. Denis J. Dean, denis.dean 'at' utdallas.edu)\n\n\tNOTE: Reuse of this database is unlimited with retention of copyright notice for Jock A. Blackard and Colorado State University.\n\n (b) Donors of database:\n\tJock A. Blackard (jblackard 'at' fs.fed.us)\n\tGIS Coordinator\n\tUSFS - Forest Inventory & Analysis\n\tRocky Mountain Research Station\n\t507 25th Street\n\tOgden, UT 84401\n\n\tDr. Denis J. Dean (denis.dean 'at' utdallas.edu)\n\tProfessor\n\tProgram in Geography and Geospatial Sciences\n\tSchool of Economic, Political and Policy Sciences\n\t800 West Campbell Rd\n\tRichardson, TX 75080-3021 \n\t\t\n\tDr. Charles W. Anderson (anderson 'at' cs.colostate.edu)\n\tAssociate Professor\n\tDepartment of Computer Science\n\tColorado State University\n\tFort Collins, CO 80523 USA\n\n (c) Date donated: August 1998\n\n3. Past Usage:\n - Blackard, Jock A. and Denis J. Dean. 2000. \"Comparative Accuracies of Artificial Neural Networks and Discriminant Analysis in Predicting Forest Cover Types from Cartographic Variables.\" Computers and Electronics in Agriculture 24(3):131-151.\n - Blackard, Jock A. and Denis J. Dean. 1998. \"Comparative Accuracies of Neural Networks and Discriminant Analysis in Predicting Forest Cover Types from Cartographic Variables.\" Second Southern Forestry GIS Conference. University of Georgia. Athens, GA. Pages 189-199.\n - Blackard, Jock A. 1998. \"Comparison of Neural Networks and\tDiscriminant Analysis in Predicting Forest Cover Types.\" Ph.D. dissertation. Department of Forest Sciences. Colorado State University. Fort Collins, Colorado. 165 pages.\n\n### Abstract of dissertation:\nNatural resource managers responsible for developing ecosystem management strategies require basic descriptive information including inventory data for forested lands to support their decision-making processes. However, managers generally do not have this type of data for inholdings or neighboring lands that are outside their immediate jurisdiction. One method of obtaining this information is through the use of predictive models.\n\nTwo predictive models were examined in this study, a feedforward neural network model and a more traditional statistical model based on discriminant analysis. The overall objectives of this research were to first construct these two predictive models, and second to compare and evaluate their respective classification accuracies when predicting forest cover types in undisturbed forests. \n\nThe study area included four wilderness areas found in the Roosevelt National Forest of northern Colorado. A total of twelve cartographic measures were utilized as independent variables in the predictive models, while seven major forest cover types were used as dependent variables. Several subsets of these variables were examined to determine the best overall predictive model. \n\nFor each subset of cartographic variables examined in this study, relative classification accuracies indicate the neural network approach outperformed the traditional discriminant analysis method in predicting forest cover types. The final neural network model had a higher absolute classification accuracy (70.58%) than the final corresponding linear discriminant analysis model(58.38%). In support of these classification results, thirty additional networks with randomly selected initial weights were derived. From these networks, the overall mean absolute classification accuracy for the neural network method was 70.52%, with a 95% confidence interval of 70.26% to 70.80%. Consequently, natural resource managers may utilize an alternative method of predicting forest cover types that is both superior to the traditional statistical methods and adequate to support their decision-making processes for developing ecosystem management strategies.\n\n - Classification performance\n - first 11,340 records used for training data subset\n - next 3,780 records used for validation data subset\n - last 565,892 records used for testing data subset\n - 70% Neural Network (backpropagation)\n - 58% Linear Discriminant Analysis\n\n4. Relevant Information Paragraph:\nPredicting forest cover type from cartographic variables only (no remotely sensed data). The actual forest cover type for a given observation (30 x 30 meter cell) was determined from US Forest Service (USFS) Region 2 Resource Information System (RIS) data. Independent variables were derived from data originally obtained from US Geological Survey (USGS) and USFS data. Data is in raw form (not scaled) and contains binary (0 or 1) columns of data for qualitative independent variables (wilderness areas and soil types).\n\nThis study area includes four wilderness areas located in the Roosevelt National Forest of northern Colorado. These areas represent forests with minimal human-caused disturbances, so that existing forest cover types are more a result of ecological processes rather than forest management practices.\n\nSome background information for these four wilderness areas: Neota (area 2) probably has the highest mean elevational value of the 4 wilderness areas. Rawah (area 1) and Comanche Peak (area 3) would have a lower mean elevational value, while Cache la Poudre (area 4) would have the lowest mean elevational value. \n\nAs for primary major tree species in these areas, Neota would have spruce/fir (type 1), while Rawah and Comanche Peak would probably have lodgepole pine (type 2) as their primary species, followed by spruce/fir and aspen (type 5). Cache la Poudre would tend to have Ponderosa pine (type 3), Douglas-fir (type 6), and cottonwood/willow (type 4). \n\nThe Rawah and Comanche Peak areas would tend to be more typical of the overall dataset than either the Neota or Cache la Poudre, due to their assortment of tree species and range of predictive variable values (elevation, etc.) Cache la Poudre would probably be more unique than the others, due to its relatively low elevation range and species composition. \n\n\n5. Number of instances (observations): 581,012\n\n6. Number of Attributes:\t\n12 measures, but 54 columns of data\t(10 quantitative variables, 4 binary wilderness areas and 40 binary soil type variables)\n\n7. Attribute information:\n\nGiven is the attribute name, attribute type, the measurement unit and a brief description. The forest cover type is the classification problem. The order of this listing corresponds to the order of numerals along the rows of the database.\n\n| Name | Data Type | Measurement | Description |\n|----------------------------------------|---------------|----------------------|------------------------------------------------------|\n| Elevation | quantitative | meters | Elevation in meters |\n| Aspect | quantitative | azimuth | Aspect in degrees azimuth |\n| Slope | quantitative | degrees | Slope in degrees |\n| Horizontal_Distance_To_Hydrology | quantitative | meters | Horz Dist to nearest surface water features |\n| Vertical_Distance_To_Hydrology | quantitative | meters | Vert Dist to nearest surface water features |\n| Horizontal_Distance_To_Roadways | quantitative | meters | Horz Dist to nearest roadway |\n| Hillshade_9am | quantitative | 0 to 255 index | Hillshade index at 9am, summer solstice |\n| Hillshade_Noon | quantitative | 0 to 255 index | Hillshade index at noon, summer soltice |\n| Hillshade_3pm | quantitative | 0 to 255 index | Hillshade index at 3pm, summer solstice |\n| Horizontal_Distance_To_Fire_Points | quantitative | meters | Horz Dist to nearest wildfire ignition points |\n| Wilderness_Area (4 binary columns) | qualitative | 0 (absence) or 1 (presence) | Wilderness area designation |\n| Soil_Type (40 binary columns) | qualitative | 0 (absence) or 1 (presence) | Soil Type designation |\n| Cover_Type (7 types) | integer | 1 to 7 | Forest Cover Type designation |\n\n\n### Code Designations:\n\n#### Wilderness Areas: \t\n| Number | Wilderness Area |\n|--------|----------------------------------|\n| 1 | Rawah Wilderness Area |\n| 2 | Neota Wilderness Area |\n| 3 | Comanche Peak Wilderness Area |\n| 4 | Cache la Poudre Wilderness Area |\n\n### Soil Types: \n1 to 40 : based on the USFS Ecological Landtype Units (ELUs) for this study area:\n\n| Study Code | USFS ELU Code | Description |\n|------------|---------------|-----------------------------------------------------------------|\n| 1 | 2702 | Cathedral family - Rock outcrop complex, extremely stony. |\n| 2 | 2703 | Vanet - Ratake families complex, very stony. |\n| 3 | 2704 | Haploborolis - Rock outcrop complex, rubbly. |\n| 4 | 2705 | Ratake family - Rock outcrop complex, rubbly. |\n| 5 | 2706 | Vanet family - Rock outcrop complex complex, rubbly. |\n| 6 | 2717 | Vanet - Wetmore families - Rock outcrop complex, stony. |\n| 7 | 3501 | Gothic family. |\n| 8 | 3502 | Supervisor - Limber families complex. |\n| 9 | 4201 | Troutville family, very stony. |\n| 10 | 4703 | Bullwark - Catamount families - Rock outcrop complex, rubbly. |\n| 11 | 4704 | Bullwark - Catamount families - Rock land complex, rubbly. |\n| 12 | 4744 | Legault family - Rock land complex, stony. |\n| 13 | 4758 | Catamount family - Rock land - Bullwark family complex, rubbly. |\n| 14 | 5101 | Pachic Argiborolis - Aquolis complex. |\n| 15 | 5151 | unspecified in the USFS Soil and ELU Survey. |\n| 16 | 6101 | Cryaquolis - Cryoborolis complex. |\n| 17 | 6102 | Gateview family - Cryaquolis complex. |\n| 18 | 6731 | Rogert family, very stony. |\n| 19 | 7101 | Typic Cryaquolis - Borohemists complex. |\n| 20 | 7102 | Typic Cryaquepts - Typic Cryaquolls complex. |\n| 21 | 7103 | Typic Cryaquolls - Leighcan family, till substratum complex. |\n| 22 | 7201 | Leighcan family, till substratum, extremely bouldery. |\n| 23 | 7202 | Leighcan family, till substratum - Typic Cryaquolls complex. |\n| 24 | 7700 | Leighcan family, extremely stony. |\n| 25 | 7701 | Leighcan family, warm, extremely stony. |\n| 26 | 7702 | Granile - Catamount families complex, very stony. |\n| 27 | 7709 | Leighcan family, warm - Rock outcrop complex, extremely stony. |\n| 28 | 7710 | Leighcan family - Rock outcrop complex, extremely stony. |\n| 29 | 7745 | Como - Legault families complex, extremely stony. |\n| 30 | 7746 | Como family - Rock land - Legault family complex, extremely stony.|\n| 31 | 7755 | Leighcan - Catamount families complex, extremely stony. |\n| 32 | 7756 | Catamount family - Rock outcrop - Leighcan family complex, extremely stony.|\n| 33 | 7757 | Leighcan - Catamount families - Rock outcrop complex, extremely stony.|\n| 34 | 7790 | Cryorthents - Rock land complex, extremely stony. |\n| 35 | 8703 | Cryumbrepts - Rock outcrop - Cryaquepts complex. |\n| 36 | 8707 | Bross family - Rock land - Cryumbrepts complex, extremely stony.|\n| 37 | 8708 | Rock outcrop - Cryumbrepts - Cryorthents complex, extremely stony.|\n| 38 | 8771 | Leighcan - Moran families - Cryaquolls complex, extremely stony.|\n| 39 | 8772 | Moran family - Cryorthents - Leighcan family complex, extremely stony.|\n| 40 | 8776 | Moran family - Cryorthents - Rock land complex, extremely stony.|\n\nZone\n\n| First digit: climatic zone | Second digit: geologic zones |\n|---------------------------------|---------------------------------------|\n| 1. lower montane dry | 1. alluvium |\n| 2. lower montane | 2. glacial |\n| 3. montane dry | 3. shale |\n| 4. montane | 4. sandstone |\n| 5. montane dry and montane | 5. mixed sedimentary |\n| 6. montane and subalpine | 6. unspecified in the USFS ELU Survey|\n| 7. subalpine | 7. igneous and metamorphic |\n| 8. alpine | 8. volcanic |\n\nThe third and fourth ELU digits are unique to the mapping unit and have no special meaning to the climatic or geologic zones.\n\n### Forest Cover Type Classes:\t\n| Category | Tree Species |\n|--------|-------------------|\n| 1 | Spruce/Fir |\n| 2 | Lodgepole Pine |\n| 3 | Ponderosa Pine |\n| 4 | Cottonwood/Willow |\n| 5 | Aspen |\n| 6 | Douglas-fir |\n| 7 | Krummholz |\n\n8. Basic Summary Statistics for quantitative variables only (whole dataset -- thanks to Phil Rennert for the summary values):\n\n| Name | Units | Mean | Std Dev |\n|----------------------------------------|------------------|---------|---------|\n| Elevation | meters | 2959.36 | 279.98 |\n| Aspect | azimuth | 155.65 | 111.91 |\n| Slope | degrees | 14.10 | 7.49 |\n| Horizontal_Distance_To_Hydrology | meters | 269.43 | 212.55 |\n| Vertical_Distance_To_Hydrology | meters | 46.42 | 58.30 |\n| Horizontal_Distance_To_Roadways | meters | 2350.15 | 1559.25 |\n| Hillshade_9am | 0 to 255 index | 212.15 | 26.77 |\n| Hillshade_Noon | 0 to 255 index | 223.32 | 19.77 |\n| Hillshade_3pm | 0 to 255 index | 142.53 | 38.27 |\n| Horizontal_Distance_To_Fire_Points | meters | 1980.29 | 1324.19 |\n\n\n9. Missing Attribute Values: None.\n\n10. Class distribution:\n| Category | Number of Records |\n|--------------------------------------|---------------------|\n| Spruce-Fir | 211,840 |\n| Lodgepole Pine | 283,301 |\n| Ponderosa Pine | 35,754 |\n| Cottonwood/Willow | 2,747 |\n| Aspen | 9,493 |\n| Douglas-fir | 17,367 |\n| Krummholz | 20,510 |\n| Other | 0 |\n| **Total Records** | **581,012** |\n\n--- \n\nJock A. Blackard\n08/28/1998 -- original text\n12/07/1999 -- updated mailing address, citations, background info for study area, added summary statistics.\n\n--- ", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "Description", "content": "### Objective\nYour primary objective is to explore the effectiveness of various machine learning algorithms and preprocessing techniques in classifying the forest cover types accurately. This project emphasises the significance of feature extraction, preprocessing, and their impact on model performance.\n\nYou are expected to use the knowledge that you have acquired in the course about classification algorithms, to come up with one model that you think is suited for this problem and which you decide to put ‘in production’.", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "Evaluation", "content": ">You are evaluated on your final report\n\nThe report should be about 10 pages. In the report you should describe and interpret your results (include lots of graphs and plots). Larger groups are expected to do more work than smaller groups.", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "rules", "content": "- Don't cheat!", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "abstract", "content": "In this project, you will tackle the Forest Cover Types dataset. The dataset includes a collection of attributes related to the Roosevelt National Forest in Colorado. There are both categorical and continuous variables, such as elevation, slope, aspect, soil type, and wilderness area. The goal of the project is to predict the forest cover type (one of seven classes) based on these features.", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "Report and Structure", "content": "1. **Introduction**\nBriefly introduce the Forest Cover Types dataset and its relevance. Highlight the classification problem of predicting forest cover types.\n\n2. **Feature Extraction and Preprocessing**\nWhat preprocessing steps have you done and what features did you use? Emphasize the pivotal role of feature extraction and preprocessing in enhancing model performance. Also discuss about feature scaling and why it is mandatory for algorithms such as k-NN.\n\n3. **Imbalanced Data Consideration**\nAcknowledge and address potential class imbalances in the dataset. Suggested techniques like oversampling, undersampling, or using synthetic data to mitigate this issue. Note that these techniques may not be suitable for this problem and therefore not necessary.\n\n4. **Algorithm Exploration**\nImplement different machine learning algorithms introduced in the course or not and see how well they compare in this problem. Remember that you must do your own training and test split.\n - If you picked any algorithm that was not described in the course, you must describe this briefly.\n - Even if you do not implement the algorithm and only use some library you still have to explain the algorithm if it was not covered in the course.\n\n5. **Hyperparameter Tuning**\nEncourage experimentation with hyperparameter tuning to optimize algorithm performance by exploring techniques such as grid search or randomized search for parameter optimization. Recall that these techniques use cross validation to select the hyperparameters. Explain why this is a good approach.\n\n6. **Evaluation Metrics**\nEmphasize the importance of selecting appropriate evaluation metrics for classification tasks. Suggest metrics like accuracy, precision, recall, and F1-score for a comprehensive assessment and model comparison.\n\n7. **Conclusion**\nThis project focuses not only on machine learning algorithms but also on the crucial steps of feature extraction and preprocessing. By comparing multiple algorithms and trying various preprocessing techniques, you will gain valuable insights into the interplay between data preparation and model performance. This approach aims to provide a deeper understanding of the nuances involved in real-world machine learning tasks.", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "Kaggle Leaderboard", "content": "There will be an optional Kaggle leader board where we test your predictions against a secret dataset. This is to help you to see how well your models are doing compared with others in the class. \n\nIn the [`Data`](https://www.kaggle.com/competitions/1dl034-project-vt24/data) section, there is a secondary dataset [`evaluation_dataset.csv`](https://www.kaggle.com/competitions/1dl034-project-vt24/data?select=evaluation_dataset.csv). This dataset does not contain any forest type classes.\nEach row has a unique ID. Your task is to categorise this dataset using your trained, tuned model and predict the forest type for each row.\nOnce done, prepare a CSV file with two columns, first column being the unique ID as it is and the corresponding forest cover category. The file should have 10982 entries.\n\n### Example CSV file for submission\n```\nID, Category\nnnn, N\nnnn, N\n....\n```\n\nSubmit the CSV file and it will be graded automatically.\n\n## Important!\nPerformance in the Kaggle leader board does not affect your mark of your project. The Kaggle competition is for motivation only, and your goal should not be just to score well on the Kaggle leader board. ", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "Other Information", "content": "> ##Data\nThe [training dataset](https://www.kaggle.com/competitions/1dl034-project-vt24/data?select=train_dataset.csv) and the [evaluation dataset](https://www.kaggle.com/competitions/1dl034-project-vt24/data?select=evaluation_dataset.csv) are available in the [`Data`](https://www.kaggle.com/competitions/1dl034-project-vt24/data) section\n\nEach group will be assigned an assistant is there to help with ideas, and to keep you on track.\n\nYou should approach the project as if you were designing a machine learning algorithm for the company. Your solution should be robust and not just finely tuned to work well with the Kaggle test set. To get your entries evaluated for the leader board they have to be in a particular format.", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false } ] }